Assessing branch efficiency and managerial behaviour in a large Spanish commercial bank

ABSTRACT

This paper describes a two-stage procedure used to evaluate determinants of operational efficiency and profit dimensions, as an indication of the performance of branch managers in a large Spanish commercial bank, before and after the bailout of the Spanish banking system in 2012. This procedure is based on both the non-parametric and robust non-parametric frontier methods and on partial frontier-based outlier detection (following Daraio and Simar) which enables us to detect potential outliers in the data, after which a classical non-parametric efficiency analysis is performed. In the first stage, efficiencies are obtained from annual data supplied by bank branches regarding the period 2011–2014. High scores were obtained in operational and profit efficiencies for DEA (pure technical and scale efficiencies), FDH and order-α estimators. In general, the following hierarchies of scores were obtained: DEA<FDH<order-α. In the second stage of our analysis, a censored econometric model approach was applied (following Simar and Wilson) to identify the drivers of efficiency. Finally, the implications of this study for policymakers are discussed.

RESUMEN

Este documento describe un procedimiento de dos etapas para evaluar los determinantes de la eficiencia en las dimensiones operativa y de beneficios en un gran banco comercial español, como indicadores del nivel de desempeño de los gerentes de las sucursales antes y después del rescate del Sistema Bancario español en 2012. Este procedimiento se basa en métodos de frontera no-paramétricos y no-paramétricos robustos, y en el método basado en la detección de outliers en fronteras parciales siguiendo a Daraio y Simar, lo que nos permite detectar valores atípicos potenciales en los datos, y después realizar un análisis clásico de eficiencia no paramétrica. En la primera etapa, las eficiencias se obtienen a partir de los datos anuales de las sucursales bancarias en el periodo 2011-2014. Los resultados indican que las puntuaciones de eficiencia operativas y de ganancias son elevadas para DEA (eficiencia técnica pura y de escala), FDH y order-α; obteniéndose la jerarquía siguiente: DEA<FDH<order-α. En la segunda etapa de nuestro análisis, se aplicó un modelo econométrico censurado siguiendo el procedimiento de Simar y Wilson para identificar los factores explicativos de la ineficiencia. Finalmente, se discuten las implicaciones de este estudio.

KEYWORDS:

• Bank branches
• technical and scale efficiencies
• DEA
• FDH
• Order-α
• Simar-Wilson procedure

JEL CLASSIFICATION:

• C02
• C44
• C61
• D24

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The non-parametric frontier approach using DEA or FDH requires minimal assumptions regarding the structure of the production and does not impose restrictions on the functional form relating inputs and outputs. Because it does not take into account noise in the data, it implicitly assumes that every deviation from the frontier can be considered as inefficiency. However, in parametric SFA, assumptions are made both about the functional form and about the distribution of the two types of error: an idiosyncratic error term and the inefficiency error term (i.e. the deviation from the frontier).

2. These authors examined commercial banks, savings banks and credit unions and calculated both cost and revenue efficiency. Their results indicated that commercial banks were more efficient than savings banks and credit unions. However, during the crisis years, the differences between these institutions shrank dramatically, especially as regards cost efficiency.

3. See the McKinsey Retail Banking Multichannel Survey, 2016; McKinsey & Company.

4. Of the 80 papers published, the largest number focuses on bank branches in Canada, followed by those on Greece, Portugal, USA and UK, respectively.

5. The same authors (p. 69) indicated that from a management perspective, the literature has reported two main types of environmental factors: region-specific (local economic growth, local unemployment rate, local community types, different opening hours, among others) and corporate-specific variables (different areas of specialisation, such as loans, mortgages, deposits, gold trading and foreign exchanges).

6. It is noteworthy that under constant returns to scale, input and output-oriented models are equivalent. However, under variable returns to scale, they are different concepts, and can lead to different efficiency measures being determined for inefficient branches.

7. If non-decreasing returns to scale (NDRS) are considered, the convexity restriction is replaced by $\sum _{j=1}^n{\lambda }_j\ge 1$.

8. This seems an attractive property of FDH since it is frequently difficult to find a good theoretical or empirical justification for postulating convex production sets in efficiency analysis. FDH is sometimes economically more meaningful than convex monotone hull, under non-trivial alternative economic conditions. Hence, FDH technical efficiency measures remain meaningful for theories of the firm that allow for imperfect competition or uncertainty.

9. The main difference from order-m is that order-α does not require resampling, and so the estimation process is faster.

10. It is among the core of strong banks that are well managed, sufficiently capitalised and apparently resilient to further shocks. Other, more vulnerable, institutions with large proportions of their balance sheets linked to the real estate sector require further restructuring and recapitalisation.

11. The gross margin is measured as the sum of the OPOs and the interest margin. OPOs are defined as the charges imposed by the bank for the provision of all types of services (total commissions + exchange differences + rest of financial transactions). The interest margin is the difference between the interest charged for credit and financial investments made (Assets) less the interest paid to customers and other lenders for the money deposited in the bank (Liabilities).

12. We also calculated output-oriented efficiency, but the results are not shown, for the sake of brevity. These results are available on request from the authors.

13. They are mainly region-specific, corporate-specific and other branch-specific factors beyond the manager’s control.

14. The number of years a company has been in existence can be a good indicator of its experience or accumulation of experience (Bernini & Guizzardi, 2010).

15. Loan loss impairment is used by some authors as input in a profit approach (Giokas, 2008a), in the view that this factor could reduce profit efficiency. However, we use the expected loan loss impairment which is related to a risk measure for the branch. More specifically, if there is a high probability that the loan will become delinquent, it is considered an expected loss. No provisions are made in this respect.

16. These categories are defined by the bank itself.

17. Paradi and Zhu (2013) pointed out that a large number of branches are located at the efficient frontier, typically 25–50%, as bank branches, in general, tend to be very well managed (due to the strict policies imposed by central offices and the government regulator).