Bank performance benchmarking and evaluation have become critically important in the dynamic banking environment in India in order to ensure sustained profitability and avoid undue risks. The Reserve Bank of India (RBI)

has taken several important steps to this end, particularly with the implementation of the recommendations of the Basel Committee for Banking Supervision, the Basel II norms. Currently, the RBI is in the process of implementing the revised Basel III norms, and is moving toward a system of Risk-Based Supervision (RBS) for monitoring bank performance. There are several systems used for bank performance evaluation. The CAMEL model, originally used by U.S. regulators to determine when to conduct on-site examination of banks, is one of the most widely-used frameworks for bank performance evaluation (Sahajwala & van der Bergh, 2000). The five CAMEL parameters including. Capital Adequacy, Asset Quality, Management

Soundness, Earnings Performance, and Liquidity Position, are essential for the sustainability of banks - insufficiency in any parameter would result in increased likelihood of bank failure.

There is extensive literature addressing bank performance evaluation. The CAMEL framework is a commonly-used methodology for bank performance assessment, using particular financial ratios to reflect different aspects of a bank’s performance (Sahajwala & van der Bergh, 2000). Several studies used the CAMEL framework as a basis for comparison of bank performance, as well as bank group performance. Dash and Das (2013) compared the performance of public sector banks with private/foreign banks under the CAMELS framework. They found that private/foreign banks fared better than public sector banks on most of the CAMELS factors in the study period, and that the primary contributing factors for the better performance of private/foreign banks were Management Soundness, and Earnings and Profitability. Several multi-criteria decision methods had been applied widely in banking performance measurement, including the Analytic Hierarchy Process (AHP), and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). Some of the literature closely linked with the present study is reviewed in the following. Several studies have applied AHP models to measure bank performance. Hunjak and Jacovcevi (2001) suggested the use of multi-criteria AHP, using both quantitative factors (viz. financial ratios) and qualitative factors (internal and external) in the evaluation process. Seçme, Bayrakdarogh, and Kahraman (2009) proposed a fuzzy AHP model for the banking system using both financial and non-financial performance criteria. Stankevičiene and Mencaitė (2012) used the AHP model to evaluate the performance of Lithuanian commercial banks. Bayyurt (2013) compared the performance of the foreign and domestic deposit banks in Turkey using several MCDM methods, namely the DEA, TOPSIS, and ELECTRE III, using the Mann-Whitney U-test and the independent samples t-test. Önder and Hepşen (2013) proposed a performance evaluation model for Turkish banks using time-series forecasting methods and multi- criteria AHP and TOPSIS methodology.

Several other multi-criteria methods discussed in the literature include ELECTRE (Bayyurt, 2013), PROMETHEE (Mareschal & Brans, 1991; Mareschal & Mertens, 1992; Babic, Belak & Tomic-Plazibat, 1999; Kosmidou & Zopounidis, 2008; Doumpos & Zopounidis, 2011; Ginevičius & Podviezko, 2013), disaggregation techniques (Zopounidis, Despotis, & Stavropoulou, 1995; Spathis, Kosanidou, & Duompos, 2002), co-plot method (Raveh, 2000), grey relational analysis (Ho, 2006), classification techniques (Ioannidis, Pasiouras, & Zopounidis, 2010), balanced scorecard approach (Wua, Tzeng, & Chen, 2009), COPRAS (Ginevičius & Podviezko, 2013), and many others. The present study attempted to integrate two approaches in bank performance measurement: multi-criteria decision models and multivariate statistical methods. The multi-criteria decision modelling approach focuses on ranking the banks according to the CAMEL parameters. The multivariate statistical approach, particularly factor analysis, is used to obtain appropriate weights for the different parameters in the CAMEL model. The scores so obtained are used to rank banks under the CAMEL dimensions, and compare the performance of public sector and private sector banks in India.

The objective of the present study was to integrate multivariate statistical methods with multi-criteria decision models in bank performance measurement. The study employed two different multi-criteria decision models to analyse bank performance: AHP and TOPSIS. The weights for the different parameters in the CAMEL model were obtained by factor analysis. The scores under the different models were in turn used to rank the sample banks, and compared the performance of public sector and private sector banks in India. The variables used in the analysis pertained to the financial ratios corresponding to the CAMEL parameters. These are discussed in the following (Dash & Das, 2013). Capital Adequacy embodies the capability of a bank in terms of sufficient capital to shield against unexpected losses. It is required in order to maintain depositors’ confidence and to prevent bankruptcy. In the current study, it was measured through three ratios: the Debt- Equity ratio, the Coverage ratio, and the Capital Adequacy ratio. Asset Quality represents the nature of loans and advances the bank has made to generate interest income. Highly rated companies generally tend to be offered lower interest rate terms than lower rated, doubtful companies. Thus asset quality mirrors the type of debtors of the bank. The ratio used to capture this parameter in this study was Net NPA to Total Advances ratio. Management Soundness is the parameter used to evaluate management quality, assigning higher value to better-managed banks and lower value to poorly-managed banks. It reflects the efficiency of management in generating business (top-line) and in maximising profits (bottom-line). In this study, it was measured via four ratios, namely. Total Investments to Total Assets ratio,

Total Advances to Total Deposits ratio, Business per Employee, and Profit per Employee. Earnings Performance focuses on how a bank earns profits. This in turn describes the sustainability and growth in earnings in the future. In this study, it was measured through three ratios, namely Return on Net Worth, Interest Spread to Total Assets ratio, and Profit after Tax to Total Assets. Liquidity Position is of prime importance in the banking business. In this study, it was measured using two ratios: Government Securities to Total Investment and Government Securities to Total Assets. The data used for the study were related to a sample of 35 banks operating in India, of which 19 were public sector banks, and 16 were private sector banks, as listed in Table 1. The research period for the study was 2007-2011. The data for the study consisted of financial ratios based on the CAMEL framework described previously , obtained from the Capitaline database1.

The weights for the different parameters in the CAMEL model, obtained by factor analysis, are presented in Table 2.

The results of the study showed that, though the AHP and TOPSIS models have resulted in a different ranking of the sample banks, there was overall consistency between the rankings resulting from the models. In particular, banks that were found to be consistently ranked high by both models can be taken as the best performers, and banks that were found to be consistently ranked low by both models can be taken as the worst performers. This would enable regulators and policy makers, on the one hand, to benchmark the performance of banks against that of best performers, and on the other hand, to take steps to improve the performance of worst performers. The AHP and TOPSIS models were found to effectively distinguish the performance of private sector banks from that of public sector banks. The results of the study need to be examined more carefully to identify the critical performance parameters for banks. The results of the study seemed to have under-rated the performance of larger banks such as SBI, HDFC Bank, and ICICI Bank. However, this is consistent with many other studies, for example, Gupta (2014) analysed 26 public sector banks using the CAMEL model in the period 2009-13, and found SBI to persistently fall in the bottom 30th percentile of public sector banks in all five dimensions. Similarly, Prasad, Ravindes and Reddy (2011) analysed 26 public sector banks and 13 private sector banks using the CAMEL model in the period 2007-11, and found SBI near the bottom 20th-25th percentile in terms of capital adequacy, asset quality, management soundness, and earnings potential, and top 25th percentile in terms of liquidity; they found HDFC Bank near the median in terms of capital adequacy, asset quality, and management soundness, in the top 5th percentile in terms of earnings potential, and top 30th percentile in terms of liquidity; and they found ICICI Bank in the bottom 10th percentile in terms of capital adequacy and asset quality, in the top 5th percentile in terms of management soundness, near the median in terms of earnings potential, and in the top 40th percentile in terms of liquidity. The difference between the perceived performance of the large banks and their scores as found in the current study could be due to the parameters considered and the analytical techniques employed. Many studies defined performance in very narrow terms, often in terms of profitability only, or using measures such as market share, customer base, number of branches, number of employees, number of ATMs, and so on, which are indicators of important aspects of banking, such as outreach and customer service, but which are highly positively correlated with size. Some analytical techniques also tend to have a size bias; for example, efficiency studies using data envelopment analysis (DEA) tend to over-rate larger banks. The CAMEL model was adopted for the study, because it is widely accepted as a comprehensive measurement model for banking performance (Sahajwala & van de Bergh, 2000). Further more, financial ratios provide better indicators of relative performance than absolute financial figures, as they avoid the size bias, and using several different disaggregating ratios for each dimension tends to give more reliable measures.

There were several limitations inherent in the current study. The study was based on a sample of only 35 banks, over only a five-year period (2007- 11), which was the period of the global financial crisis, which seriously affected banking systems worldwide. Thus, the results of the study may be period specific, and may not be generalisable. Also, the approach followed in the current study considered only some performance parameters, and did not consider some qualitative aspects of banking performance, such as management performance and staff efficiency. Further more, the study did not analyse the sensitivity of any the scores to the parameter weights. Also, the study used factor analysis, which determined weights in order to maximise the variance explained, but which may not reflect the importance of the parameters in banking performance. There is a vast scope for further research in the area of bank performance and risk measurement, particularly as the banking sector is highly dynamic in nature. Several other multi-criteria models may be used to analyse banking performance to provide alternative perspectives to regulators and policy makers, for example, ELECTRE and PROMETHEE methodologies can be used to identify banks that dominate over other banks in terms of performance, DEA methodology may be used to identify inefficient banks, and VIKOR methodology may be used to identify critical trade-offs in banking performance.