Literature review of bank efficiency around the world

Efficiency studies in finance and banking via DEA are also voluminous in recent years.

Ayadi et al. (1998) attempted to determine the quality of bank management in Nigeria by

using DEA approach from 1991 to 1994. Karim (2001) investigated the cost efficiency of 155 banks from five ASEAN countries during the period from 1989 to 1996. Krivonozhko et al. (2002) made the efficiency analysis of 150 Russian banks in 1998, just after the banking default in Russia in the same year. Pastor and Perez (2002) analyzed cost and profit efficiency of banks in ten countries of the European Union during the period from 1993 to 1996. Jemric and Vujcic (2002) evaluated bank efficiency in Croatia from 1996 to 2000 by using the DEA approach. Sathye (2003) measured the productive efficiency of banks in India during the period from 1997 to 1998. Ozkan-Gunay and Tektas (2006) measured the technical efficiency of Turkish non-public banks from 1990 to 2001 via CCR DEA model.

Based upon the decomposition technical efficiency from pure technical efficiency and scale efficiency, Aly et al. (1990) utilized a nonparametric frontier approach to compute the overall, technical, allocative, and scale efficiency for a sample of 322 U.S. independent banks in 1986. They took real estate loans, commercial and industrial loans, consumer loans, all other loans and demand deposits as five outputs, labor, capital and loanable funds as three inputs and three price information of each input. Their empirical results indicated that their sample banks are characterized by relatively low level of overall efficiency, and these banks could produced the same level of output by using only 65% of the inputs actually used.

Thus, inefficiency in these banks might be attributed to under-utilization or wasting of inputs.

For the advanced analysis, the pooled sample banks were split into two sub-samples: banks that are allowed to operate branches (212) and those that are prohibited from operating branches (110) to test the null hypothesis that the two sub-sample banks were drawn from the same population (environment). Their null hypothesis could not be rejected in their study, which means the two sub-sample banks would face the same environment. Finally, they used multiple regression analysis to conclude that the diversity of financial products and bank

location are significantly accounted for the inefficiency.

Huang (1997) conducted a translog cost function with three inputs (deposits, labor and capital) and three outputs (financial investment, short-term loans and long-term loans) to examine the cost efficiency of twenty-two banks in Taiwan from 1981 to 1992. In 1992, Taiwan regulatory government allowed the establishment of new stock-shared bank and its branches. Huang almost described clearly the process of bank evolution and the business operation of Taiwan’s banks, and his work is almost a complete review before the Taiwan’s financial opening in 1992. The input-output variable specification and translog cost function in his model are almost followed by later studies.

Seiford and Zhu (1999) examined the performance of the top fifty-five U.S. commercial banks via two production process which separates profitability and marketability. Their two-stage production is profit earning in the first stage and market value generating in the second stage. Their sample banks were drawn from the Fortune 1000 (Fortune April 29, 1996), ranking by revenue. Their procedure is divided into two stages and eight factors are expressed as inputs and outputs in each stage. The first stage measured profitability, i.e., a bank’s ability to generate the revenue and profit in terms of its labor, assets and capital in financial market. The second stage measures marketability, i.e. a bank generates market value, total returns to investors, and earnings per share in the stock market by profit and revenue. It can be seen that profit and revenue serve as mediator in bank production that they are the outputs from the first stage and the inputs to the second stage. The efficiency of both two stages is based on CCR DEA model. Their empirical findings indicated that close to 90% of the banks are inefficient in both profitability and marketability. Furthermore, most large banks exhibit better performance on profitability, whereas smaller banks tend to perform

better with respect to marketability. This suggested that bank size may have a positive effect on profitability but negative effect on marketability.

With respect to the extended application of DEA measurement, Kao and Liu (2004) attempted to predict the performance of twenty-four commercial banks in Taiwan via CCR DEA model. They claimed the insufficiency of prediction only by using financial ratios such as ROE, EPS, P/E ratio, etc. The prediction of the sample banks is based on their own financial forecasts disclosed by themselves. The uncertain forecasts of financial data are presented in ranges (lower and upper bound), so the results of the prediction of the efficiency scores which are computed by interval data are also in ranges. According to the comparison of the results form their prediction and traditional financial ratios, a bank that had several financial poor ratios last year could obtain higher predicted efficiency scores next year.

Therefore, the authors suggested that it is misleading to only use financial ratio analysis to examine, evaluate, and even to forecast the performance of a bank, especially in the financial environment with rapid changes and intense competition.

Wang et al. (2005) used nonparametric DEA models, including CCR, BCC, Bilateral, Slack-Based Measure and the FDH models, to evaluate the overall, pure and scale efficiencies of the sixteen nationwide commercial banks in mainland China. The sixteen banks in China are classified into two ownership groups: four state-owned banks and twelve stock-shared banks. The authors used total capital and total assets as the two input variables, and net profit, return on equity (ROE) and return on assets (ROA) as the three output variables.

Their study concluded with three findings below: First, the FDH model analysis can not distinguish between efficient and inefficient banks from their sample banks; nevertheless, the CCR and BCC models can do. Second, seven banks is in the increasing returns-to-scale

stage, which means that those banks can improve their performance by increasing their size, and the others are in the decreasing returns-to-scale stage. Third, on average, the stock-shared banks have higher efficiency than the state-owned banks do. However, the authors only took year 2004 as study period. The neglect of panel data might cause incomplete results. They either did not consider the environmental factors which influence the performance directly and indirectly that it is not adequate to define the reason of inefficiency.

Cheng and Dran (2005) used stochastic frontier model to estimate the efficiency of nineteen major banks in China from 1998 to 2002. With respect to the effect of exogenous factors on bank’s inefficiency, they concluded that joining WTO is a positive force to improve the efficiency of China banking industry, however, the duration of establishment, the amount of total loans and the state-owned ownership are negative factors to account for the efficiency.

It is suggested that the ownership reform to stock-shared type and decrease of total loans can improve the efficiency under the current circumstance.

Howland and Rowse (2006) used the DEA model to measure the efficiency of branches of a major Canadian bank to compare with the results of a U.S. bank studied by Golany and Storbeck (1999) during the same period. With almost the same input and output variables, the U.S. bank has higher averaged efficiency score; however, the score distribution of the Canadian bank is more central (with lower standard deviation of efficiency scores). The reason may be the sole regulatory supervision and banking regulation in Canada. The authors asserted that Canadian banks are more homogenous rather than U.S. banks.

Hu et al. (2006) investigated the efficiency of twelve nationwide banks in China from

1996 to 2003. They used standard CRS and VRS DEA model to make the cost efficiency analysis and seemingly unrelated regression model to examine the relation of external operational environment and banks’ performance. Their empirical findings indicate as follows: First, nationwide joint-equity commercial banks have significantly higher cost, overall technical, and scale efficiencies, but lower pure technical efficiency than state-owned specialized banks. The next, a marginal increasing relation exists between the deposit-loan ratio and cost efficiency and an inverted U-shape relation exists between the deposit-loan ratio and overall technical as well as scale efficiencies. Third, small-sized banks have higher cost and allocative efficiencies than large-sized banks do. Fourth, the twelve banks have lower cost efficiency after the 1997 Asian financial crisis and 2001 WTO participation, and they have lower overall technical, pure technical, and scale efficiencies after 2001. Finally, the twelve sample banks have significantly increasing overall technical and scale efficiencies from 1996 to 2003.

With respect to an attempt to further discussion on the relation of banks’ efficiency and their stock returns, Kirkwood and Nahm (2006) surveyed the cost efficiency of Austrian banks during the period from 1995 to 2002. Since the Austrian banking is dominated by four banks (the major banks), the authors utilized the VRS DEA model to compare the performance of the major banks and regional banks. The authors used two datasets with the same input components (employee, net fixed assets and interest-bearing liabilities) but the different output components (interest-bearing assets and non-interest income versus profit before tax) to investigate the banking service efficiency and their findings indicated that the major banks have better profit but poorer banking service efficiency. Based upon their computation of profit efficiency scores, the authors then used multiple regression, taking the excess return on stock (return on stock minus the risk free rate) as independent variable and

taking excess market return and percentage change in profit efficiency as dependent variables, to test the significance of efficiency on stock return. Efficiency changes of banks are fully reflected in stock returns.