## _cr07199 - References

## Source details

**Canonical URL:** [_cr07199 - References](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2007/_cr07199.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2007/_cr07199.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2007/_cr07199.pdf.json)

---

### Overview
- Prepared in the context of the 2006 Switzerland FSAP update; analysis of structure, medium-term trends, key challenges of the Swiss banking system; includes assessment of various efficiency indicators based on data envelopment models and policy recommendations.
- Banking system segmentation described: large internationally-oriented banks (UBS and Credit Swiss), three domestically-oriented retail segments (cantonal, regional, Raiffeisen), foreign banks, Swiss private banks, and assorted institutions.
- Key summary findings:
  - Estimated cost inefficiencies account for about 15–25 percent of costs.
  - Domestically-oriented retail segments systematically under-perform in cost efficiency relative to large and private banks.
  - Domestically-oriented segments operate under increasing returns to scale (sizes below optimal efficient scale).
  - Widening efficiency gap between benchmark institutions and less efficient performers over the last ten years.
  - Rebound in total factor productivity growth since 2002 after relative stagnation between 1995 and 2001.

### System size, composition, and trends
- System-scale and macro contribution:
  - Total system assets accounted for roughly six-times GDP in 2005.
  - Deposits in custody accounts held off-balance sheet represented almost ten times GDP in 2005.
  - Banking contributes about 12 percent of GDP and employs three percent of the labor force.
  - Average labor productivity in banking is about four times overall productivity in the economy.
- Dualistic nature and segmentation:
  - Two large banks (UBS and Credit Swiss) increased share of system assets from about one-half in 1995 to two-thirds in 2005.
  - Smaller institutions are fragmented and generally domestically oriented; foreign and Swiss private banks tend to focus on private banking for non-residents.
- Consolidation and scale dynamics:
  - Number of institutions: from 495 in 1990 to 404 in 1996 and to 337 in 2005.
  - Number of branches decreased by 21 percent from 4,424 in 1996 to 3,503 in 2005.
  - Aggregated bank balance sheet increased from about three-times GDP in 1996 to six-times GDP in 2005.
  - Average bank size rose from ChF 3.7 billion in 1996 to ChF 8.4 billion in 2005.
  - Segmental asset share changes: Raiffeisen share rose from 2.9 percent in 1999 to 3.8 percent in 2005; regional and cantonal banks’ joint share fell from 25.3 percent in 1995 to 14.4 percent in 2005.

### Market structure and activity composition
- Concentration measures (1995 to 2005):
  - Herfindhal concentration index increased from 1,020 in 1995 to 2,690 in 2005.
  - Five-bank concentration ratio increased from 60 percent in 1995 to 76 percent in 2005.
  - Herfindhal index in 2005 by segment: cantonal banks 1,040; regional banks 950.
- Shifts in activity composition:
  - Trading portfolios in large banks increased from 14.3 percent in 2001 to 20.6 percent in 2005.
  - Securities maintained off-balance sheet in custody accounts increased by 51.9 percent in 2002–2005.
  - Reliance on customer deposits as financing increased across all segments since 2000.

### Revenue and performance indicators
- Revenue composition and trends:
  - Interest revenues accounted for 66 percent of total revenues in 2005, down from 71 percent in 1996.
  - Large banks (2005): interests ~70 percent, fee income ~20 percent, trading income ~10 percent of total revenues.
  - Foreign banks (2005): fee income 44 percent of total revenues.
  - Private banks (2005): fee income 75 percent of total revenues.
  - Cantonal, regional, and Raiffeisen banks depend almost exclusively on interest revenues, although declining over time.
- Customary performance statistics (system and comparators):
  - Operating costs to assets fell from 2 percent in 2000 to 1½ percent in 2005.
  - Cost-to-income ratios increased from 53.7 percent in 2000 to 63 percent in 2005.
  - Return on assets (ROA) increased from 0.5 percent in 2002 to 0.9 percent in 2005; comparator-country average ROA 0.6 percent.
  - Comparator snapshot (2005):
    - Switzerland — Assets to GDP 623, Credit to GDP 389, 5-bank Concentration Ratio 76, Herfindahl Index 2,690, Number of Credit Institutions 337, Number of branches per Th. inhabitants 0.5.
    - Comparator-country averages: Average Assets to GDP 499, Credit to GDP 165, 5-bank Concentration Ratio 52, Herfindahl Index 932, Number of Credit Institutions 325, Number of branches per Th. inhabitants 0.4.

### Efficiency assessment methodology and inputs/outputs
- Methodology:
  - Data envelopment analysis (DEA) and Malmquist total factor productivity indexes applied to estimate cost efficiency and productivity evolution.
  - DEA constructs an efficient frontier; two distance orientations by duality used:
    - Input-orientated: minimizes inputs given outputs (static, cross-section).
    - Output-orientated: maximizes outputs given inputs (time-evolution, panel).
- Efficiency concepts estimated:
  - Technical efficiency (TE): TE = 1 – ab/ao = bo/ao (value between 0 and 1).
  - Allocative efficiency (AE): example AE = co/bo.
  - Cost efficiency (CE): CE = AE × TE.
  - Returns to scale (RTS): IRS, CRS, DRS inferred by comparing VRS, NIRS, CRS TE estimates.
- Input-output specifications (intermediation approach):
  - Model 1 (preferred): Outputs = (i) loans, (ii) trading securities, (iii) securities in custody accounts; Inputs = deposits, personnel, number of branches. Input prices: average interest expenses = total interest expenses / deposits; average personnel remuneration = total personnel expenses / number of staff; cost of branches = other operating costs (excluding interest, personnel expenses, and loan provisions) / number of branches. Fixed assets not considered (0.6 percent of total bank assets, Switzerland, end-2005).
  - Model 2 (alternative): Outputs = (i) interest revenues; (ii) non-interest revenues (net trading income + commission and fee income); Inputs = deposits, personnel, number of branches.
- Robustness and validity:
  - Two validity conditions: comparable technologies across banks; inputs/outputs reflect banking activity.
  - Four complementary exercises:
    1. Baseline: entire population classified in seven segments.
    2. Domestically-oriented retail sample: cantonal, regional, Raiffeisen banks.
    3. International comparison: domestically-oriented retail segments versus France, Germany, Spain, U.K.
    4. Time evolution: Malmquist index 1996–2005 with Fare et al. (1994) decomposition.

### Baseline efficiency results (Model 1, end-2005)
- System averages:
  - Average cost efficiency for the entire system: 81.7 percent (implying average gap of 18.3 percent).
- Benchmark segments (cost efficiency score 100):
  - Large banks: 100.0 (CRS and VRS)
  - Private banks: 100.0 (CRS and VRS)
- Segment scores (Model 1, exact values):
  - CRS panel:
    - Cantonal banks: Technical Efficiency 85.3, Allocative Efficiency 82.3, Cost Efficiency 70.2
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 74.0, Cost Efficiency 74.0
    - Large banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0
    - Other banks: Technical Efficiency 73.1, Allocative Efficiency 78.9, Cost Efficiency 57.7
    - Private banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0
    - Raiffeisen banks: Technical Efficiency 77.7, Allocative Efficiency 79.1, Cost Efficiency 61.5
    - Regional banks: Technical Efficiency 84.2, Allocative Efficiency 78.7, Cost Efficiency 66.3
    - Mean: Technical Efficiency 88.6, Allocative Efficiency 84.7, Cost Efficiency 55.7
  - VRS panel:
    - Cantonal banks: Technical Efficiency 88.8, Allocative Efficiency 85.2, Cost Efficiency 75.7 (IRS)
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 74.9, Cost Efficiency 74.9 (-)
    - Large banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0 (-)
    - Other banks: Technical Efficiency 75.8, Allocative Efficiency 76.7, Cost Efficiency 58.1 (DRS)
    - Private banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0 (-)
    - Raiffeisen banks: Technical Efficiency 89.7, Allocative Efficiency 85.0, Cost Efficiency 76.3 (IRS)
    - Regional banks: Technical Efficiency 100.0, Allocative Efficiency 89.7, Cost Efficiency 89.7 (IRS)
    - Mean: Technical Efficiency 93.5, Allocative Efficiency 87.4, Cost Efficiency 82.1
- Key interpretations:
  - Average efficiency relatively high; technical efficiency slightly better than allocative efficiency.
  - Domestically-oriented retail segments show excessive personnel and number of branches relative to outputs and operate under increasing returns to scale (IRS).

### Medium-term evolution (1996–2005), Model 1
- Mean VRS changes (2001–2005 relative to 1996–2000):
  - Mean (Average 2001-2005): Technical Efficiency 96.4, Allocative Efficiency 87.9, Cost Efficiency 84.9
  - Mean (Difference 2001-05 relative to 1996-00): Technical Efficiency -3.6, Allocative Efficiency -7.1, Cost Efficiency -10.1
- CRS panel representative changes:
  - Mean (Average 2001-2005): Technical Efficiency 90.8, Allocative Efficiency 85.7, Cost Efficiency 78.2
  - Mean (Difference 2001-05 relative to 1996-00): Technical Efficiency -9.2, Allocative Efficiency -9.3, Cost Efficiency -16.8
- Segment-specific VRS changes:
  - Cantonal banks: VRS cost efficiency change -9.0 percentage points.
  - Raiffeisen banks: VRS cost efficiency change -7.6 percentage points.
  - Regional banks: VRS cost efficiency change 1.8 percentage points (narrowed gap due to improved allocative efficiency).
- Robustness (Model 2, alternative outputs) overall averages:
  - CRS mean: Technical Efficiency 79.3, Allocative Efficiency 84.8, Cost Efficiency 68.4
  - VRS mean: Technical Efficiency 92.4, Allocative Efficiency 90.1, Cost Efficiency 83.9
  - Overall average efficiency score reported: 83.9 percent.
- Model 2 segment highlights (exact values):
  - CRS:
    - Cantonal banks: Technical Efficiency 60.4, Allocative Efficiency 81.6, Cost Efficiency 49.3
    - Raiffeisen banks: Technical Efficiency 52.4, Allocative Efficiency 78.3, Cost Efficiency 41.0
    - Regional banks: Technical Efficiency 59.3, Allocative Efficiency 78.7, Cost Efficiency 46.7
    - Mean: Technical Efficiency 79.3, Allocative Efficiency 84.8, Cost Efficiency 68.4
  - VRS:
    - Cantonal banks: Technical Efficiency 68.2, Allocative Efficiency 83.3, Cost Efficiency 56.9 (IRS)
    - Raiffeisen banks: Technical Efficiency 78.6, Allocative Efficiency 79.5, Cost Efficiency 62.5 (IRS)
    - Regional banks: Technical Efficiency 100.0, Allocative Efficiency 77.0, Cost Efficiency 77.0 (IRS)
    - Mean: Technical Efficiency 92.4, Allocative Efficiency 90.1, Cost Efficiency 83.9
- Interpretation:
  - Model 2 yields similar qualitative ordering but generally lower efficiency estimates for domestically-oriented retail segments; cantonal, Raiffeisen, and regional show IRS under Model 2.

### Domestically-oriented retail banks: focused assessment (VRS, Model 1, 2005)
- Sample: 104 institutions (24 cantonal banks, 79 regional banks, Raiffeisen treated as single entity). Represent:
  - 18 percent of assets
  - 66 percent of total number of branches
  - 23 percent of total bank employees
- Summary efficiency estimates (VRS, unweighted averages, standard deviations in parenthesis):
  - Cantonal banks:
    - Technical Efficiency 90.7 (10.3)
    - Allocative Efficiency 92.0 (6.5)
    - Cost Efficiency 83.8 (13.3)
    - Number of Banks 24; Number of Benchmark Banks 6
  - Raiffeisen banks:
    - Technical Efficiency 100.0 (n.a.)
    - Allocative Efficiency 100.0 (n.a.)
    - Cost Efficiency 100.0 (n.a.)
    - Number of Banks 1; Number of Benchmark Banks 1
  - Regional banks:
    - Technical Efficiency 87.7 (12.9)
    - Allocative Efficiency 88.5 (6.6)
    - Cost Efficiency 77.6 (13.3)
    - Number of Banks 79; Number of Benchmark Banks 8
  - Entire Sample:
    - Technical Efficiency 88.5 (12.4)
    - Allocative Efficiency 89.4 (6.7)
    - Cost Efficiency 79.2 (13.6)
    - Number of Banks 104; Number of Benchmark Banks 15
- Interpretations:
  - Average cost efficiency for the sample: 79.2 percent.
  - Technical and allocative efficiency similar—input mix roughly cost-minimizing.
  - Cost efficiency not systematically related to bank size (quartile distribution shows similar averages across quartiles).
  - Benchmark institutions distributed across size quartiles: out of 15 benchmarks, 4 in first quartile, 7 in fourth quartile.

### Cross-country perspective (France, Germany, Spain, U.K.)
- Dataset: OECD data covering five years 1999–2003; resident banks including domestic and foreign branches and domestic subsidiaries (foreign subsidiaries excluded).
- Model 1 (outputs = loans and securities holdings; inputs = deposits, personnel, number of branches) and Model 2 (outputs = interest revenues and non-interest revenues; same inputs).
- Currency conversion: OECD financial data converted into US dollars using end-of-period market exchange rates.
- Cross-country results (Model 1 summary):
  - Overall cost efficiency ranges for Swiss segments: cantonal 71.8 percent, regional 95.6 percent.
  - Swiss cantonal, regional, and cooperatives exceed sample average cost efficiency (entire sample mean 65.5 percent).
  - Swiss scores higher than cooperative banks in France, Germany, Spain and savings banks in France and Spain.
  - Differences largely attributable to technical efficiency; Swiss banks perform relatively well in allocative efficiency.
  - Model suggests the three Swiss segments are operating under IRS.
- Robustness (Model 2):
  - Similar qualitative ranking but generally lower efficiency estimates for Swiss segments.
  - Cantonal banks overall cost-efficiency under Model 2: 43.6 percent (implying gap of 56.4 percent with comparator institutions).
  - Background calculations suggest cantonal banks’ volumes of deposits and number of branches too large for their interest and non-interest revenues.
  - Conclusion: cantonal banks somewhat less efficient relative to peer samples in this check.

### Time evolution of productivity: Malmquist index results (1996–2005)
- DEA specification for Malmquist exercise:
  - Outputs: interest revenues and non-interest revenues.
  - Inputs: deposits, personnel, number of branches.
- System-wide yearly averages (Malmquist Index and components):
  - Malmquist Index (year-on-year):
    - 1997: 1.152
    - 1998: 1.032
    - 1999: 1.052
    - 2000: 0.998
    - 2001: 0.959
    - 2002: 1.007
    - 2003: 1.082
    - 2004: 1.061
    - 2005: 1.066
    - Mean: 1.044
  - Technical Efficiency (CRS) (year-on-year):
    - 1997: 0.970
    - 1998: 0.985
    - 1999: 0.974
    - 2000: 1.018
    - 2001: 1.003
    - 2002: 1.039
    - 2003: 1.010
    - 2004: 0.997
    - 2005: 0.956
    - Mean: 0.994
  - Technical Change (year-on-year):
    - 1997: 1.188
    - 1998: 1.048
    - 1999: 1.080
    - 2000: 0.980
    - 2001: 0.956
    - 2002: 0.969
    - 2003: 1.071
    - 2004: 1.065
    - 2005: 1.115
    - Mean: 1.050
- Aggregate interpretation:
  - Average improvement in total factor productivity: 4.4 percent per year (100×(1.044–1)).
  - Technical change main driver: average increase of 5 percent per year.
  - Technical efficiency decreased on average 0.6 percent per year (100×(0.994–1) = −0.6 percent).
  - Productivity stagnated between 1999 and 2002, picked up strongly since 2002, mainly induced by technical change since 2003.
- Segment-level Malmquist averages (1997–2005):
  - Cantonal banks: Malmquist Index 1.042; Technical Efficiency (CRS) 0.990; Technical Change 1.052.
  - Foreign banks: Malmquist Index 1.022; Technical Efficiency (CRS) 0.975; Technical Change 1.048.
  - Large banks: Malmquist Index 1.069; Technical Efficiency (CRS) 1.000; Technical Change 1.069.
  - Other banks: Malmquist Index 1.069; Technical Efficiency (CRS) 1.003; Technical Change 1.065.
  - Private banks: Malmquist Index 1.058; Technical Efficiency (CRS) 1.000; Technical Change 1.058.
  - Raiffeisen banks: Malmquist Index 1.010; Technical Efficiency (CRS) 0.985; Technical Change 1.025.
  - Regional banks: Malmquist Index 1.041; Technical Efficiency (CRS) 1.007; Technical Change 1.034.
  - Mean (system): Malmquist Index 1.044; Technical Efficiency (CRS) 0.994; Technical Change 1.050
- Segment interpretations:
  - Largest productivity improvements: Large banks and Other banks (Malmquist Index 1.069 each); Private banks 1.058.
  - Large banks and private banks: average technical efficiency 1.000 (operating at the frontier).
  - Regional banks and other banks: evidence of catching-up (technical efficiency > 1.000).
  - Foreign banks, cantonal, and Raiffeisen: technical efficiency < 1.000 (widening distance to benchmark for some segments).
- Dynamics:
  - Cumulated Malmquist index shows productivity growth in large banks consistently above system average; gap increased after 2003.
  - Productivity growth in cantonal and regional banks roughly similar and catching up with system average.
  - Raiffeisen banks: productivity improving since 2000 at a lower pace.

### Policy implications and recommendations
- Consolidation outlook:
  - A possible deceleration of potential growth could renew consolidation pressures; cost inefficiencies and economies of scale may be key drivers of consolidation.
  - Continued strong domestic competition and limited growth opportunities likely to foster further cost-sharing arrangements, exits, and mergers among smaller institutions.
  - From a stability perspective, further consolidation among relatively small institutions is unlikely to entail systemic risks if exits occur sparsely over time.
- Cantonal banks and public ownership considerations:
  - Ensure cantonal banks operate with sufficient flexibility to conduct business consistent with profit maximization; resources used to pay for social ends should ideally be taken exclusively from taxes and dividends paid by the cantonal banks to their cantonal governments.
  - Strengthen governance structures of cantonal banks to shield operations from unwarranted political interference on business objectives and managerial decisions.
  - Consider medium-term government divestment from the banking sector.
- Four main conclusions:
  1. Domestically-oriented retail bank segments have potential to improve further their cost efficiency.
  2. Performance of these segments compares relatively well against similar segments in the four countries included in the study.
  3. Large banks and private banks are systematically selected as benchmarks; efficiency gap with other segments has been increasing over the last ten years.
  4. Overall productivity in the Swiss banking sector remained relatively constant during 1996–2001, but has picked up strongly since 2002.

*Source: SNB and IMF staff calculations; Prepared by Francisco Vázquez (Monetary and Capital Markets Department).*

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

### _cr07199 - References

### Overview
- Prepared in the context of the 2006 Switzerland FSAP update; analysis of structure, medium-term trends, key challenges of the Swiss banking system; includes assessment of various efficiency indicators based on data envelopment models and policy recommendations.
- Banking system segmentation described: large internationally-oriented banks (UBS and Credit Swiss), three domestically-oriented retail segments (cantonal, regional, Raiffeisen), foreign banks, Swiss private banks, and assorted institutions.
- Key summary findings:
  - Estimated cost inefficiencies account for about 15–25 percent of costs.
  - Domestically-oriented retail segments systematically under-perform in cost efficiency relative to large and private banks.
  - Domestically-oriented segments operate under increasing returns to scale (sizes below optimal efficient scale).
  - Widening efficiency gap between benchmark institutions and less efficient performers over the last ten years.
  - Rebound in total factor productivity growth since 2002 after relative stagnation between 1995 and 2001.

### System size, composition, and trends
- System-scale and macro contribution:
  - Total system assets accounted for roughly six-times GDP in 2005.
  - Deposits in custody accounts held off-balance sheet represented almost ten times GDP in 2005.
  - Banking contributes about 12 percent of GDP and employs three percent of the labor force.
  - Average labor productivity in banking is about four times overall productivity in the economy.
- Dualistic nature and segmentation:
  - Two large banks (UBS and Credit Swiss) increased share of system assets from about one-half in 1995 to two-thirds in 2005.
  - Smaller institutions are fragmented and generally domestically oriented; foreign and Swiss private banks tend to focus on private banking for non-residents.
- Consolidation and scale dynamics:
  - Number of institutions: from 495 in 1990 to 404 in 1996 and to 337 in 2005.
  - Number of branches decreased by 21 percent from 4,424 in 1996 to 3,503 in 2005.
  - Aggregated bank balance sheet increased from about three-times GDP in 1996 to six-times GDP in 2005.
  - Average bank size rose from ChF 3.7 billion in 1996 to ChF 8.4 billion in 2005.
  - Segmental asset share changes: Raiffeisen share rose from 2.9 percent in 1999 to 3.8 percent in 2005; regional and cantonal banks’ joint share fell from 25.3 percent in 1995 to 14.4 percent in 2005.

### Market structure and activity composition
- Concentration measures (1995 to 2005):
  - Herfindhal concentration index increased from 1,020 in 1995 to 2,690 in 2005.
  - Five-bank concentration ratio increased from 60 percent in 1995 to 76 percent in 2005.
  - Herfindhal index in 2005 by segment: cantonal banks 1,040; regional banks 950.
- Shifts in activity composition:
  - Trading portfolios in large banks increased from 14.3 percent in 2001 to 20.6 percent in 2005.
  - Securities maintained off-balance sheet in custody accounts increased by 51.9 percent in 2002–2005.
  - Reliance on customer deposits as financing increased across all segments since 2000.

### Revenue and performance indicators
- Revenue composition and trends:
  - Interest revenues accounted for 66 percent of total revenues in 2005, down from 71 percent in 1996.
  - Large banks (2005): interests ~70 percent, fee income ~20 percent, trading income ~10 percent of total revenues.
  - Foreign banks (2005): fee income 44 percent of total revenues.
  - Private banks (2005): fee income 75 percent of total revenues.
  - Cantonal, regional, and Raiffeisen banks depend almost exclusively on interest revenues, although declining over time.
- Customary performance statistics (system and comparators):
  - Operating costs to assets fell from 2 percent in 2000 to 1½ percent in 2005.
  - Cost-to-income ratios increased from 53.7 percent in 2000 to 63 percent in 2005.
  - Return on assets (ROA) increased from 0.5 percent in 2002 to 0.9 percent in 2005; comparator-country average ROA 0.6 percent.
  - Table 1 comparator snapshot (2005): Switzerland — Assets to GDP 623, Credit to GDP 389, 5-bank Concentration Ratio 76, Herfindahl Index 2,690, Number of Credit Institutions 337, Number of branches per Th. inhabitants 0.5.
  - Comparator-country averages listed: Average Assets to GDP 499, Credit to GDP 165, 5-bank Concentration Ratio 52, Herfindahl Index 932, Number of Credit Institutions 325, Number of branches per Th. inhabitants 0.4.

### Efficiency assessment methodology and results (high-level)
- Methodology: Data envelopment analysis (DEA) and Malmquist total factor productivity indexes applied to estimate cost efficiency and productivity evolution.
- Key efficiency results:
  - About 15–25 percent of costs attributed to inefficiencies (point estimates).
  - Large banks and private banks are consistently selected as benchmark (more efficient) performers in various exercises.
  - Domestically-oriented banks (cantonal, regional, Raiffeisen) show room for efficiency gains and operate under increasing returns to scale.
  - Time-series findings indicate widening gaps between benchmarks and less efficient banks over ten years.
  - Productivity trends show a rebound in total factor productivity growth since 2002 following stagnation between 1995 and 2001.

### Policy implications and recommendations
- Consolidation outlook:
  - A possible deceleration of potential growth could renew consolidation pressures; cost inefficiencies and economies of scale may be key drivers of consolidation.
  - Continued strong domestic competition and limited growth opportunities likely to foster further cost-sharing arrangements, exits, and mergers among smaller institutions.
  - From a stability perspective, further consolidation among relatively small institutions is unlikely to entail systemic risks if exits occur sparsely over time.
- Recommendations regarding cantonal banks:
  - Ensure cantonal banks operate with sufficient flexibility to conduct business consistent with profit maximization; resources used to pay for social ends should ideally be taken exclusively from taxes and dividends paid by the cantonal banks to their cantonal governments.
  - Strengthen governance structures of cantonal banks to shield operations from unwarranted political interference on business objectives and managerial decisions.
  - Consider medium-term government divestment from the banking sector.

*Prepared by Francisco Vázquez (Monetary and Capital Markets Department); content from the IMF technical note prepared in the context of the 2006 Switzerland FSAP update.*

### 18.      As in other industries, assessing economic performance in banking is difficult.

### _cr07199 - 18.      As in other industries, assessing economic performance in banking is difficult.

### Assessment challenges and common indicators
- Commonly-used financial indicators such as ROA, or cost-to-income ratios, convey useful information on financial performance, but offer little guidance to identify best practices and are clearly insufficient to assess operating efficiency.
- Cost-to-income ratios: computed by dividing operating expenses (excluding interest expenses and loan-loss provisions), by the sum of net-interest revenues plus non-interest income.
- Profits per employee: doubled from ChF 92.6 thousands per employee in 2002 to ChF 207.5 thousands in 2005.
- Table 2 (Indicators of Bank Performance in Comparator Countries, 2005) highlights selected aggregate ratios. Examples reported:
  - Average: Net interest income to assets 1.3; Total expenses to assets 1.3; ROA 0.65; Interest income to total income 58.6; Cost-to-income ratio 58.7.
  - Switzerland: Net interest income to assets 0.8; Total expenses to assets 1.5; ROA 0.9; Interest income to total income 60.9; Cost-to-income ratio 63.0.

### Data envelopment analysis (DEA) approach — purpose and orientations
- DEA is used to analyze cost efficiency in the Swiss banking system. DEA: a linear programming technique for performance evaluation and benchmarking in a multivariate setting.
- DEA constructs an efficient frontier enveloping the data and uses it as a benchmark to assess efficiency indicators for individual entities.
- Two distance orientations by duality:
  - Input-orientated: minimizes the use of inputs given a target output mix — used for static, cross-section performance assessment.
  - Output-orientated: maximizes output using a given input mix — used to assess time evolution of performance and efficiency indicators (panel dimension).

### Efficiency concepts estimated with DEA
- Technical efficiency (TE): ability to minimize input use given a target output mix (or equivalently to maximize output with given inputs). Illustrative expression: TE = 1 – ab/ao = bo/ao (value between 0 and 1; 1 indicates on the efficient isoquant).
- Allocative efficiency (AE): ability to select the cost-minimizing input combination given input prices and a target output mix. Example relative measure AE = co/bo.
- Cost efficiency (CE): compounded effect (product) of technical and allocative efficiency; provides a measure of the gap between an observed input-output mix and its cost-minimizing alternative. Expressed as CE = AE × TE.
- Returns to scale (RTS): measure of the change in production in response to a proportional increase in all inputs; used to compare observed bank size with efficient scale.
- Scale effects and returns-to-scale assumptions:
  - CRS assumption valid only when banks operate at efficient scale. TE estimates can be biased by scale efficiencies (SE).
  - TE can be computed under NIRS and VRS to relax CRS and to obtain scale effect estimates.
  - Three possible combinations (as described): relationships among VRS, NIRS, and CRS TE estimates indicate IRS, CRS, or DRS cases.

### Implementation choices, robustness exercises, and limitations
- Two validity conditions for DEA results:
  1. Technologies used by banks in each estimation must be comparable (institutions dedicated to similar activities).
  2. Sets of inputs and outputs must adequately reflect the nature of banking activity.
- Swiss context: universal banking status supports computing efficiency for the entire population of banks, but specialization across banks raises caveats.
- Four complementary exercises to address methodological limitations and robustness:
  1. Baseline: efficiency estimates using the entire population of Swiss banks classified in seven segments (large banks, foreign banks, private banks, cantonal banks, cooperative banks (Raiffeisen), regional banks, other banks).
  2. Domestically-oriented retail sample: restrict to cantonal, regional, and Raiffeisen banks to account for comparable technologies.
  3. International comparison: efficiency for domestically-oriented retail segments in Switzerland versus comparable segments in France, Germany, Spain, and the UK.
  4. Time evolution: Malmquist index of total factor productivity growth for 1996–2005 with Fare et al. (1994) decomposition (changes in technical efficiency and changes in technology).

### Input-output specifications used
- Intermediation approach (Sealey and Lindley, 1977) guided input/output choices. Two alternative model specifications used to check robustness:
  - Model 1 (preferred): three outputs and three inputs.
    - Outputs: (i) loans, (ii) trading securities, (iii) securities in custody accounts. Note: loans and trading securities represent 85 percent of bank assets in 2005.
    - Inputs: deposits, personnel, number of bank branches. Input prices estimated as:
      - average interest expenses = total interest expenses / deposits;
      - average personnel remuneration = total personnel expenses / number of staff;
      - cost of branches = other operating costs (excluding interest, personnel expenses, and loan provisions) / number of branches.
    - Fixed assets not considered because they represent 0.6 percent of total bank assets (Switzerland, end-2005).
    - Two variants explored with roughly similar results: (i) two-output using loans and trading securities; (ii) two-output using loans and total securities — both neglect securities in custody accounts.
  - Model 2 (alternative): two outputs and same inputs.
    - Outputs: (i) interest revenues; (ii) non-interest revenues (sum of net trading income, plus commission and fee income). This specification encompasses entire operating revenues and expenses.

### Malmquist index and productivity decomposition
- Output-orientated Malmquist index used to measure productivity evolution and to disentangle:
  - changes in technical efficiency (catching up),
  - changes in technology (innovation).
- The Malmquist index combines four distance measures across two periods to separate the efficiency-change component (catch-up) and the technical-change component (frontier evolution).
- Under this metric:
  - Index > 1 implies productivity improvement.
  - Index < 1 implies productivity deterioration.
  - Components individually > 1 indicate improvement in that component; < 1 indicate deterioration.

### Data and exercises summary
- Data: bank-level dataset comprising the population of Swiss banks, classified by bank segments, covering 1995–2005.
- Four exercises exploited the dataset:
  - Cross-segment comparison across the entire population of banks (baseline).
  - Comparison among domestically-oriented retail banks: cantonal, regional, Raiffeisen.
  - Cross-country cost efficiency comparison for domestically-oriented retail segments in France, Germany, Spain, and the UK.
  - Time evolution of productivity and efficiency indicators across seven Swiss bank segments during 1996–2005.
- Estimations executed using code developed by Coelli (1996). Data provided by the Swiss National Bank (SNB).

*Source: ECB, SNB and IMF staff calculations; IMF staff note using DEA and Malmquist methods as described in the text.*

### 34.      The baseline exercise assessed efficiency across the seven bank segments in

### _cr07199 - 34.      The baseline exercise assessed efficiency across the seven bank segments in

### Baseline efficiency results (aggregate, end-2005, Model 1)
- Data: end-2005, aggregated by bank segments; all efficiency scores in percent.  
- Interpretation: score below 100 indicates failure to operate at the efficient frontier (unexploited efficiency gains).  
- Specification comparison:
  - CRS (upper panel) generally yields lower efficiency estimates than VRS (lower panel) because VRS fits the frontier better.
- System averages (Model 1):
  - Average cost efficiency for the entire system: 81.7 percent (implying average gap of 18.3 percent, i.e., 100–81.7).
- Benchmark segments (overall cost efficiency score of 100):
  - Large banks: 100.0 (CRS and VRS)
  - Private banks: 100.0 (CRS and VRS)
- Foreign banks:
  - Technical efficiency (VRS): 100.0
  - Allocative efficiency (VRS): 72.1
  - Cost efficiency (CRS): 74.0; (VRS): 74.9
- Segment scores, Model 1 (2005) — exact values from Table 3:
  - Constant Returns to Scale:
    - Cantonal banks: Technical Efficiency 85.3, Allocative Efficiency 82.3, Cost Efficiency 70.2
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 74.0, Cost Efficiency 74.0
    - Large banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0
    - Other banks: Technical Efficiency 73.1, Allocative Efficiency 78.9, Cost Efficiency 57.7
    - Private banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0
    - Raiffeisen banks: Technical Efficiency 77.7, Allocative Efficiency 79.1, Cost Efficiency 61.5
    - Regional banks: Technical Efficiency 84.2, Allocative Efficiency 78.7, Cost Efficiency 66.3
    - Mean: Technical Efficiency 88.6, Allocative Efficiency 84.7, Cost Efficiency 55.7
  - Variable Returns to Scale:
    - Cantonal banks: Technical Efficiency 88.8, Allocative Efficiency 85.2, Cost Efficiency 75.7 (IRS)
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 74.9, Cost Efficiency 74.9 (-)
    - Large banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0 (-)
    - Other banks: Technical Efficiency 75.8, Allocative Efficiency 76.7, Cost Efficiency 58.1 (DRS)
    - Private banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0 (-)
    - Raiffeisen banks: Technical Efficiency 89.7, Allocative Efficiency 85.0, Cost Efficiency 76.3 (IRS)
    - Regional banks: Technical Efficiency 100.0, Allocative Efficiency 89.7, Cost Efficiency 89.7 (IRS)
    - Mean: Technical Efficiency 93.5, Allocative Efficiency 87.4, Cost Efficiency 82.1
- DEA specification: three outputs (loans and trading securities, and securities in custody accounts) and three inputs (deposits, personnel, and number of branches).

### Key interpretation and segment-specific findings
- Overall: average efficiency relatively high; technical efficiency slightly better than allocative efficiency.
- Domestically-oriented retail segments (cantonal, Raiffeisen, regional, other banks):
  - Lowest efficiency scores under both CRS and VRS: “other banks.”
  - Cantonal, Raiffeisen, and regional: roughly similar cost-efficiency estimates; ordering sensitive to model specification.
  - Intermediate calculations (not in table) indicate excessive personnel and number of branches relative to outputs for these three segments.
  - Evidence of scale inefficiency:
    - TE estimates under VRS larger than under CRS for these segments.
    - NIRS TE estimates (not shown) indicate these segments operate under increasing returns to scale (IRS) — sizes too small to exploit scale economies.

### Medium-term evolution (1996–2005), Model 1 (Table 4)
- Method: data split into 1996–2000 and 2001–2005; efficiency scores computed for each sub-period; last three columns in Table 4 show change (2001–05 relative to 1996–00); negative value = drop.
- Main outcomes:
  - Average VRS mean widened by 10.1 percentage points (i.e., mean cost-efficiency change under VRS: -10.1).
  - Cantonal banks: VRS cost efficiency change -9.0 percentage points.
  - Raiffeisen banks: VRS cost efficiency change -7.6 percentage points.
  - Regional banks: VRS cost efficiency change 1.8 percentage points (narrowed gap due to improved allocative efficiency).
- Selected exact values (VRS panel, Average 2001–2005 and Difference 2001-05 relative to 1996-00):
  - Mean (Average 2001-2005): Technical Efficiency 96.4, Allocative Efficiency 87.9, Cost Efficiency 84.9
  - Mean (Difference 2001-05 relative to 1996-00): Technical Efficiency -3.6, Allocative Efficiency -7.1, Cost Efficiency -10.1
- CRS panel (representative changes):
  - Mean (Average 2001-2005): Technical Efficiency 90.8, Allocative Efficiency 85.7, Cost Efficiency 78.2
  - Mean (Difference 2001-05 relative to 1996-00): Technical Efficiency -9.2, Allocative Efficiency -9.3, Cost Efficiency -16.8

### Robustness check: Model 2 (2005) — alternative output definition
- Model 2 outputs: interest revenues and trading revenues plus fee income; inputs: deposits, personnel, number of branches.
- Overall averages:
  - CRS mean: Technical Efficiency 79.3, Allocative Efficiency 84.8, Cost Efficiency 68.4
  - VRS mean: Technical Efficiency 92.4, Allocative Efficiency 90.1, Cost Efficiency 83.9
  - Overall average efficiency score reported: 83.9 percent (text).
- Segment highlights (Table 5 exact values):
  - CRS:
    - Cantonal banks: Technical Efficiency 60.4, Allocative Efficiency 81.6, Cost Efficiency 49.3
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 78.9, Cost Efficiency 78.9
    - Large banks: 100.0 / 100.0 / 100.0
    - Other banks: Technical Efficiency 83.2, Allocative Efficiency 76.0, Cost Efficiency 63.2
    - Private banks: 100.0 / 100.0 / 100.0
    - Raiffeisen banks: Technical Efficiency 52.4, Allocative Efficiency 78.3, Cost Efficiency 41.0
    - Regional banks: Technical Efficiency 59.3, Allocative Efficiency 78.7, Cost Efficiency 46.7
    - Mean: Technical Efficiency 79.3, Allocative Efficiency 84.8, Cost Efficiency 68.4
  - VRS:
    - Cantonal banks: Technical Efficiency 68.2, Allocative Efficiency 83.3, Cost Efficiency 56.9 (IRS)
    - Foreign banks: Technical Efficiency 100.0, Allocative Efficiency 100.0, Cost Efficiency 100.0 (-)
    - Large banks: 100.0 / 100.0 / 100.0 (-)
    - Other banks: Technical Efficiency 100.0, Allocative Efficiency 91.0, Cost Efficiency 91.0 (DRS)
    - Private banks: 100.0 / 100.0 / 100.0 (-)
    - Raiffeisen banks: Technical Efficiency 78.6, Allocative Efficiency 79.5, Cost Efficiency 62.5 (IRS)
    - Regional banks: Technical Efficiency 100.0, Allocative Efficiency 77.0, Cost Efficiency 77.0 (IRS)
    - Mean: Technical Efficiency 92.4, Allocative Efficiency 90.1, Cost Efficiency 83.9
- Interpretation:
  - Model 2 yields similar qualitative ordering but generally lower efficiency estimates for domestically-oriented retail segments (notably cantonal, Raiffeisen, regional).
  - Under Model 2, foreign banks included in benchmark set.
  - Cantonal, Raiffeisen, regional show IRS under Model 2; difference between VRS TE and CRS TE for these three larger than under Model 1.

### Limitations noted
- Potential differences in business technologies between bank segments may affect comparability; differences in business focus can lead to technology differences. The next section examines this in more detail.

### Domestically-oriented retail banks: focused assessment (Section B)
- Sample: 104 institutions out of 337 system-wide, covering:
  - 24 cantonal banks
  - 79 regional banks
  - Raiffeisen banks (treated as single entity for pooled cooperatives)
- These institutions represent:
  - 18 percent of assets
  - 66 percent of total number of branches in the country
  - 23 percent of total bank employees
- Implication: combined sample has more branches and more employees per unit of assets than system average.
- Summary efficiency estimates for domestically-oriented retail banks (Table 6, VRS, Model 1, 2005; unweighted averages, standard deviations in parenthesis):
  - Cantonal banks:
    - Technical Efficiency 90.7 (10.3)
    - Allocative Efficiency 92.0 (6.5)
    - Cost Efficiency 83.8 (13.3)
    - Number of Banks 24
    - Number of Benchmark Banks 6
  - Raiffeisen banks:
    - Technical Efficiency 100.0 (n.a.)
    - Allocative Efficiency 100.0 (n.a.)
    - Cost Efficiency 100.0 (n.a.)
    - Number of Banks 1
    - Number of Benchmark Banks 1
  - Regional banks:
    - Technical Efficiency 87.7 (12.9)
    - Allocative Efficiency 88.5 (6.6)
    - Cost Efficiency 77.6 (13.3)
    - Number of Banks 79
    - Number of Benchmark Banks 8
  - Entire Sample:
    - Technical Efficiency 88.5 (12.4)
    - Allocative Efficiency 89.4 (6.7)
    - Cost Efficiency 79.2 (13.6)
    - Number of Banks 104
    - Number of Benchmark Banks 15
- Interpretations:
  - Average cost efficiency for the entire sample: 79.2 percent.
  - Technical and allocative efficiency are similar — input mix roughly consistent with cost minimization.
  - Relative ordering differs from prior exercises: cantonal banks above regional banks in both technical and allocative efficiency here.
  - Average gaps: technical efficiency gap 3 percent (90.7–87.7); allocative efficiency gap 3.5 percent.
- Size relationship:
  - Cost efficiency does not appear systematically related to bank size.
  - Table 7 (distribution by quartiles of bank size) exact excerpts:
    - Cantonal banks average cost efficiency by quartile: Q1 74.4, Q2 85.6, Q3 83.8; Number of benchmark banks in Cantonal: Q2 6; Number of banks across quartiles: 4, 2, 0, 24
    - Raiffeisen banks average cost efficiency: 100.0 in Q4; Number of benchmark banks: 1; Number of banks: 1
    - Regional banks average cost efficiency by quartile: Q1 82.8, Q2 75.8, Q3 74.0, Q4 75.5; Number of benchmark banks by quartile: Q1 4, Q2 13.8; Number of banks by quartile: 26, 26, 26, 22 (total 79)
    - Total average cost efficiency by quartile: Q1 82.8, Q2 75.8, Q3 74.1, Q4 84.2; Total mean 79.2
    - Total number of benchmark banks by quartile: Q1 4, Q2 13, Q3 7, Q4 1; Total number of banks by quartile: 26, 26, 26, 26 (total 104)
  - Observed: first-quartile banks average cost efficiency 82.8 percent; fourth-quartile 82.4 percent — close values.
  - Benchmark institutions distribution: out of 15 benchmark institutions, 4 in first quartile, 7 in fourth quartile.
- Figure 3 (not reproduced here): supports that distribution of cost efficiency for domestically-oriented banks is not related to size.  
- DEA specification for this domestically-oriented exercise: three outputs (loans, trading securities, securities in custody accounts) and three inputs (deposits, personnel, number of bank branches).

### Cross-country perspective (Section C)
- Countries compared: France, Germany, Spain, U.K.; Swiss segments compared: cantonal banks, Raiffeisen, regional and savings banks.
- OECD data: five years 1999–2003; resident banks including domestic and foreign branches and domestic subsidiaries; foreign subsidiaries excluded.
- Segment mapping varied by country (Table 8 provides list of sampled bank segments by country).
- Two model specifications used:
  - Model 1: outputs = loans and securities holdings; inputs = deposits, personnel, number of branches. Input prices approximated by: interest expenses / customer deposits; personnel expenses / number employees; operating expenses (net of interest and personnel expenses) / number of branches. (OECD dataset limitation: securities holdings not separable into trading and non-trading.)
  - Model 2: outputs = interest revenues and non-interest revenues; same inputs; parallel computation of input prices.
- Currency conversion: OECD financial data converted into US dollars using end-of-period market exchange rates.
- Cross-country results (Table 9 summary in text):
  - Under Model 1 (left panel):
    - Overall cost efficiency ranges for Swiss segments: cantonal 71.8 percent, regional 95.6 percent.
    - Swiss cantonal, regional, and cooperatives exceed sample average cost efficiency (entire sample mean 65.5 percent).
    - Swiss scores higher than cooperative banks in France, Germany, Spain and savings banks in France and Spain.
    - Differences largely attributable to technical efficiency; Swiss banks also perform relatively well in allocative efficiency.
    - Model suggests the three Swiss segments are operating under IRS.
  - Robustness (Model 2, right panel):
    - Similar qualitative ranking but generally lower efficiency estimates for Swiss segments.
    - Cantonal banks overall cost-efficiency under Model 2: 43.6 percent (implying gap of 56.4 percent with comparator institutions).
    - Background calculations (not shown) suggest cantonal banks’ volumes of deposits and number of branches too large for their interest and non-interest revenues.
    - Conclusion: cantonal banks somewhat less efficient relative to peer samples in this cross-country check; contrasts with Rime and Stiroh (2003), but thorough reconciliation requires further statistical exploration.

*Source: SNB and IMF staff calculations.*

### 48. The time evolution of productivity in Switzerland was assessed with a Malmquist

### _cr07199 - 48. The time evolution of productivity in Switzerland was assessed with a Malmquist

### Methodology
- Productivity evolution assessed with an output-orientated Malmquist productivity index using yearly data during 1996–2005.
- Based on the methodology described in section II-B: constructs a best-practice technology frontier against which individual bank segments are compared.
- Estimation required the distances specified in equation (6), entailing 196 optimization problems.
- Yearly evolution of the Malmquist index and its two components was computed for each bank segment; yearly averages taken over all bank segments are summarized in Table 10.
- DEA exercise specification (used for results reported):
  - Two outputs: interest revenues and non-interest revenues.
  - Three inputs: deposits, personnel, and number of branches.
- Malmquist index decomposition (as in Appendix A):
  - Malmquist index = catch-up (change in relative efficiency) × technical change (evolution of production frontier).
  - Distance function D_t(x_t, y_t) defined; D_t = 1 if observed input-output lies on frontier, otherwise D_t < 1.

### System-wide results (Table 10; yearly averages over all bank segments)
- Year-on-year Change in:
  - Malmquist Index:
    - 1997: 1.152
    - 1998: 1.032
    - 1999: 1.052
    - 2000: 0.998
    - 2001: 0.959
    - 2002: 1.007
    - 2003: 1.082
    - 2004: 1.061
    - 2005: 1.066
    - Mean: 1.044
  - Technical Efficiency (CRS):
    - 1997: 0.970
    - 1998: 0.985
    - 1999: 0.974
    - 2000: 1.018
    - 2001: 1.003
    - 2002: 1.039
    - 2003: 1.010
    - 2004: 0.997
    - 2005: 0.956
    - Mean: 0.994
  - Technical Change:
    - 1997: 1.188
    - 1998: 1.048
    - 1999: 1.080
    - 2000: 0.980
    - 2001: 0.956
    - 2002: 0.969
    - 2003: 1.071
    - 2004: 1.065
    - 2005: 1.115
    - Mean: 1.050
- Aggregate interpretation:
  - Average improvement in total factor productivity was 4.4 percent per year (i.e., 100×(1.044–1)).
  - Technical change was the main driving force: average increase of 5 percent per year.
  - Technical efficiency decreased at an average 0.6 percent per year (mean TE = 0.994 implies 100×(0.994–1) = −0.6 percent).
  - Overall productivity improved since 2002, mainly induced by technical change since 2003.
  - Evidence of counteracting effects between technical change and efficiency (e.g., 2000–2002: receding technical change accompanied by catch-up in technical efficiency).

### Segment-level results (Table 11; average Malmquist indexes by bank segments, 1997–2005)
- Year-on-year Change in:
  - Cantonal banks:
    - Malmquist Index: 1.042
    - Technical Efficiency (CRS): 0.990
    - Technical Change: 1.052
  - Foreign banks:
    - Malmquist Index: 1.022
    - Technical Efficiency (CRS): 0.975
    - Technical Change: 1.048
  - Large banks:
    - Malmquist Index: 1.069
    - Technical Efficiency (CRS): 1.000
    - Technical Change: 1.069
  - Other banks:
    - Malmquist Index: 1.069
    - Technical Efficiency (CRS): 1.003
    - Technical Change: 1.065
  - Private banks:
    - Malmquist Index: 1.058
    - Technical Efficiency (CRS): 1.000
    - Technical Change: 1.058
  - Raiffeisen banks:
    - Malmquist Index: 1.010
    - Technical Efficiency (CRS): 0.985
    - Technical Change: 1.025
  - Regional banks:
    - Malmquist Index: 1.041
    - Technical Efficiency (CRS): 1.007
    - Technical Change: 1.034
  - Mean (system):
    - Malmquist Index: 1.044
    - Technical Efficiency (CRS): 0.994
    - Technical Change: 1.050
- Segment-level interpretation:
  - Largest improvement in total factor productivity:
    - Large banks: 6.9 percent per year (Malmquist Index 1.069).
    - Other banks: 6.9 percent per year (Malmquist Index 1.069).
    - Private banks: 5.8 percent per year (Malmquist Index 1.058).
  - Cantonal banks attained improvements in technology slightly higher than the system average (Technical Change 1.052 vs. mean 1.050).
  - Large banks and private banks: average technical efficiency of 1.000 → consistently operating at the frontier over the period.
  - Regional banks and other banks: evidence of catching-up (average technical efficiency > 1.000 for these segments).
  - Widening distance to benchmark observed for some segments: foreign banks, cantonal, and the Raiffeisen (technical efficiency < 1.000).

### Dynamics and cumulated patterns
- Overall average productivity stagnated between 1999 and 2002, and picked up strongly afterwards.
- Cumulated Malmquist index (sequential product of yearly indexes) shows:
  - Productivity growth in large banks consistently above system average; gap increased after 2003.
  - Productivity growth in cantonal and regional banks roughly similar and catching up with system average.
  - Raiffeisen banks: productivity improving since 2000, at a lower pace.

### Policy implications and conclusions (from V. CONCLUSIONS AND POLICY RECOMMENDATIONS)
- Swiss banking sector undergoing substantial consolidation over the last ten years, but retains varied institutional types across business segments.
- Consolidation driven partly by competitive pressures following housing market bubble burst; resulted in substantial reduction in number of institutions and branches.
- Smaller banks adopting cost-sharing arrangements and outsourcing back operations to benefit from scale economies.
- The diverse institutional landscape adds to strength but challenges medium-term profitability for domestically-oriented retail banks due to expected slow-down in economic activity induced by population aging.
- Under current benign macro environment, banks post robust results, easing immediate pressure for cost-cutting; medium-term competitive pressures expected to resume, potentially forcing new exits.
- Systemic financial stability risk from consolidation is limited due to small size of institutions, particularly if consolidation occurs over extended period.
- Public policy considerations because of public ownership of cantonal banks and contingent public liability from deposit guarantees:
  - Consider strengthening governance structures of cantonal banks to shield from unwarranted political interference.
  - Recommend clear and strong separation between social goals and business operations of cantonal banks (e.g., use cantonal banks taxes and dividends to finance legitimate social goals while avoiding interference with business decisions).
  - Over the medium term, further public divestment from the banking system could be considered.
- Four main conclusions:
  1. Domestically-oriented retail bank segments have potential to improve further their cost efficiency.
  2. Performance of these segments compares relatively well against similar segments in the four countries included in the study.
  3. Large banks and private banks are systematically selected as benchmarks; efficiency gap with other segments has been increasing over the last ten years.
  4. Overall productivity in the Swiss banking sector remained relatively constant during 1996–2001, but has picked up strongly since 2002.

*Source: SNB and IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2007/_cr07199.pdf_