Table 1.
Brief description of objectives and results of each analysed study.
| N. | Study | Brief Summary | Number of Citations |
|---|---|---|---|
| 1 | Luo (2003) | By analysing three gaps in the literature, the authors examine the profitability and marketability efficiency of 245 US banks, as well as verify whether the bank location impacts its efficiency. The results suggest that banks’ greatest source of inefficiency is marketability. The bank’s location is unrelated to efficiency ratios, and overall technical efficiency can be used as a predictor of the likelihood of bank failure. | 134 |
| 2 | Barth and Staat (2005) | The study analyses the efficiency of 31 agencies from Kölner Bank in Germany, verifying the impact of non-discretionary variables such as branch area, public transport, competition and others. The two-stage models were able to more accurately evaluate efficiency compared to one-stage models, and none of the environmental variables were statistically significant. | 9 |
| 3 | Wu et al. (2006) | The paper combines DEA with Neural Network (DEA-NN) to analyse the relative efficiency of branches of a Canadian bank. The results found by the proposed model are comparable to those of the traditional DEA models. The proposed model leads to a more robust boundary and identifies more efficient DMUs, but it is inferior in identifying benchmarks. | 162 |
| 4 | Pasiouras (2008) | The study evaluates bank efficiency in different countries (different contexts) by checking the impact of regulatory factors on efficiency. The results provide evidence for relevance of the three pillars of Basel’s II Accord. Larger banks with lower loans showed better technical efficiency indices under all circumstances. Country-specific variables had a statistically significant impact on efficiency. | 105 |
| 5 | Mostafa (2009) | The paper analyses the efficiency of the largest Arab banks through the integration of DEA with neural networks (NN). NN have great potential to assess the relative efficiency of banks because of their flexibility and robustness. The predictive capacity of the model is very similar to the results of other statistical techniques. | 51 |
| 6 | Thoraneenitiyan and Avkiran (2009) | The authors investigate the relationship between post-crisis banking restructuring and country-specific factors with bank efficiency in a sample of 110 banks in five Asian countries in the period 1997 to 2001. Bank restructuring does not necessarily increase banks’ efficiency. Domestic M&A perform better on efficiency than foreign acquisition. Banks under state intervention are more inefficient. The inefficiencies in the banking sector are attributed largely to country- specific factors. | 30 |
| 7 | Staub et al. (2010) | The study estimates cost, allocative and technical efficiencies of Brazilian banks in the post-privatization period (2000–2007) through a three-stage model. Brazilian banks have a high degree of inefficiency compared to other countries. Stated owned banks were more efficient than private ones, and foreigners showed higher levels of cost inefficiency. Size is not an important variable that impacts efficiency. | 87 |
| 8 | Tsolas (2010) | The study measures the overall efficiency, comprised by the efficiency of profitability and effectiveness, of the 50 best branches of a Greek bank. Nineteen branches were efficient in profitability and effectiveness. Regarding overall efficiency, the main cause of inefficiency was profitability. The bank’s performance can be largely improved by changing practices in branches identified as worst DMUs. | 20 |
| 9 | Azadeh et al. (2011) | The authors analyse and suggest strategies to optimize the productivity of workers from various branches of the Bank of Industry and Mining in Iran. Integrating AHP and DEA, the study verified that a large part of the inefficiency of the branches is due to low work quality level and high number of training hours. The proposed analysis technique leads to better results than others, exploring both qualitative and quantitative data. | 24 |
| 10 | Holod and Lewis (2011) | The authors propose a DEA model that considers the variable deposits as an intermediate variable. The results show that the decision to define deposits as input or output significantly affects the indexes and the efficiency ranking of traditional models and, for this reason, the method developed by the authors managed to avoid this dilemma. | 51 |
| 11 | Paradi et al. (2011) | The authors apply a two-stage DEA model to analyse the efficiency of 816 bank branches in order to reconcile the results indicated by this model with the opinions of the managers of these organizations. The efficiency indexes presented considerable variations among the different regions analysed. Branches in smaller markets were more efficient. Considering different approaches for analysing efficiency allowed finding results with greater consistency. | 111 |
| 12 | Shahroodi et al. (2011) | The study analyses the efficiency of 20 branches of Saderat Bank in Iran, pointing out which units are efficient and inefficient, as well as benchmarking inefficient ones and how they can improve their operations. Only three branches were efficient, and the largest source of inefficiency was in the production stage. | 2 |
| 13 | Liu and Chen (2012) | The authors identify bank failures through a two-stage DEA model of worst practices, which makes it possible to work with negative outputs. The empirical analysis showed the applicability of the model to predict potential bank failures. The model predicted a number of potential banks to fail similar to what has been observed in Taiwan. | 2 |
| 14 | Maghyereh and Awartani (2012) | The study analyses the influence of reforms in the banking sector of six countries, whose objectives were to strengthen the financial and economic integration between these countries. These measures had a significant impact on the efficiency and homogenization of the banking sectors of the countries analysed. | 17 |
| 15 | Shyu and Chiang (2012) | The authors analyse the true managerial efficiency of 123 branches of a bank in Taiwan, through a three-stage model, adjusted for environmental variables and statistical noise. Traditional DEA models overestimated efficiency ratios. The main cause of branch inefficiency was the operated scale. Location did not show significant impacts on efficiency. Branches with greater scope of action and volume of deposits were more efficient. | 23 |
| 16 | Yang and Liu (2012) | The study integrates NDEA with Fuzzy to measure branch performance in Taiwan’s banking industry. Most of the branches analysed had a better performance in the first stage of productivity. Interest cost is the largest factor in the first stage, while fund transfer income and interest income are key factors of the second stage. | 43 |
| 17 | Halkos and Tzeremes (2013) | The study examines the efficiency of 18 Greek banks in a period of Greece’s fiscal crisis by checking how the banks’ efficiency would react to possible Mergers and Acquisitions (M&A). The results suggest that, analysing the year before and the year after the crisis, M&A did not generate operational efficiency in the short term. M&A between efficient banks will not necessarily generate an efficient bank. | 32 |
| 18 | Kholousi (2013) | The paper measures the efficiency of 16 branches in Iran using an integrated DEA model with AHP. The location of the branches was a key factor of efficiency. The strengths of one branch can serve as benchmarking for the others. The use of AHP together with DEA provided more consistent results. | 1 |
| 19 | Lin and Chiu (2013) | The authors apply an integrated model for the measurement of bank efficiency in Taiwan through Independent Component Analysis (ICA) and the Network Slack-Based Measure (NSBM). Three dimensions of efficiency were analysed: production efficiency, service efficiency and profitability efficiency. The results indicate that the proposed model was able to determine the main causes of bank inefficiency, presenting an excellent discriminatory feature. | 23 |
| 20 | Matthews (2013) | The study evaluates the risk management performance of Chinese banks in terms of their contribution to profitability through a three-stage NDEA model. The inclusion of the proxies for risk improved the efficiency measurement. | 69 |
| 21 | Özdemir (2013) | The study integrates DEA and Analytical Network Process (ANP) to evaluate the efficiency of commercial banks in Turkey, with the possibility of incorporating managerial preferences into the model. The proposed integration presented several advantages over traditional models, such as considering multiple performance measures. The weights of the model can based on the preferences of the managers. | 1 |
| 22 | Xu (2013) | The paper verifies the impact of size and market power on the efficiency of 16 Chinese banks in the period from 2007 to 2011. The results found that size is a determinant of the efficiency of banks. A favourable economic environment (real GDP growth) also has a positive influence on efficiency. | 0 |
| 23 | Ebrahimnejad et al. (2014) | The authors propose a three-stage DEA model with two independent parallel stages, where the outputs of these stages serve as input to the third stage, with the presence of undesirable outputs. In a case study of 49 People’s Bank branches, the study corroborates the effectiveness and applicability of the model in bank efficiency studies. | 29 |
| 24 | Huang et al. (2014) | The study proposes a two-stage DEA Network Slack-Based Measures (NSBM) model with undesirable output aiming to open the black box of the production process. The proposed model has a better applicability than traditional models. All hypotheses suggested for efficiency determinants were confirmed for overall efficiency. On the other hand, the hypotheses could not be accepted when each stage were analysed individually. | 9 |
| 25 | Piot-Lepetit and Nzongang (2014) | The authors analyse the efficiency of Micro Financial Institutions (MFIs) both in the execution of financial tasks and in their role of coping with social problems, through, for instance, loans to poor people. In 46% of MFIs, there was no trade-off between the two dimensions analysed. Directives were given in order for MFIs to improve both their financial and social efficiencies. | 18 |
| 26 | Wang et al. (2014) | The study analyses the efficiency of the 16 largest Chinese banks in the period from 2003 to 2011, which corresponds to a reform in the Chinese banking sector. The authors consider deposits as intermediary variable and unrealized loans as undesirable output. The two-stage model was able to explain more appropriately the inefficiency of the banks than conventional DEA models. The efficiency of the banks increased during the period analysed because of the reform. State owned banks were more efficient before the reform, however difference to other banks decrease afterwards. | 76 |
| 27 | Wang et al. (2014) | The authors investigate the relationship between bank efficiency and intellectual capital in a sample of 16 US banks through a two-stage model. Profitability is included in the first stage and creation of value is included in the second. The authors found evidence that intellectual capital positively impacts efficiency. | 21 |
| 28 | Wanke and Barros (2014) | The study evaluates the 40 largest banks in Brazil regarding the optimization of costs and productive efficiency, establishing a connection between these two variables. Brazilian banks tend to be more efficient at translating administrative expenses and personnel expenses into shareholders’ equity and fixed assets than at managing physical and human resources. M&A, size and the fact that the bank is state owned are also variables that influence efficiency. | 48 |
| 29 | An et al. (2015) | The authors measure the efficiency of 16 Chinese banks in the period 2008 to 2012 through a two-stage DEA-SBM approach, in which the first stage was called a deposit generator and the second as a deposit user with the presence of undesirable output. The results indicate that efficiency has increased during these five years due to banks’ improvements in deposit creation. | 11 |
| 30 | Chao et al. (2015) | The study applies the Dynamic Network Slack-Based Measure Data Envelopment Analysis Model (DNSBM) to evaluate the performance of Taiwanese banks during the period 2005–2011. Using a three-stage model, the results indicate that banks have lost profitability since the 2008 crisis, while the creation of intellectual capital increased from 2008 to 2010. | 7 |
| 31 | Khalili-Damghani et al. (2015) | The paper evaluates the relative efficiency of customer services in 30 branches of an Iranian bank, through a hybrid model based on Multi-Criteria Satisfaction Analysis (MUSA) and NDEA. The proposed method was able to identify which branches were able to meet consumers’ expectations. | 0 |
| 32 | Fukuyama and Weber (2015) | The authors assess the dynamic efficiency and productivity of Japanese commercial banks, maximizing desirable outputs and minimizing undesirable outputs (Non Performing Loans). For a 3 year dynamic window, the inefficiency of Japanese banks ranged from 19.5% of average outputs and inputs in 2007–2009 to 21.5% of average outputs and inputs in 2008–2010. Banks could become more efficient by increasing the volume of deposits. | 17 |
| 33 | Kwon et al. (2015) | The authors combine two empirical data analysis techniques to evaluate and predict performance improvements for 181 US banks. The proposed model contributes, in an impactful way, to the managerial process of decision making. | 19 |
| 34 | Shawtari et al. (2015) | The authors measure the efficiency of Islamic Yemeni commercial banks, analysing the stability and efficiency of the sector. The study also checks for variables that may be affecting efficiency. The results suggest that the recent reforms adopted by the Yemeni government have failed to improve the sector since the efficiency scores were low. Islamic banks performed better than commercial banks. | 2 |
| 35 | Sufian (2015) | The study estimates the efficiency of Malaysian banks from 1999 to 2008, analysing the impact of several environmental variables such as liquidity, risk, size, profitability, capitalization level, macroeconomic conditions. Size, non-interest income, foreign control, and capitalization have a positive impact on productive efficiency. State owned banks were more inefficient. Credit risk and liquidity were not statistically significant. | 0 |
| 36 | Tsolas and Charles (2015) | The paper incorporates stochastic models in the DEA to analyse the efficiency of Greek banks in a period of crisis of the country incorporating variable related to risk. The model measures efficiency considering the possibility of stochastic variables in the DEA model, In addition, the model is able to control, from the efficiency indexes, the favourable operating conditions. | 21 |
| 37 | Wang and Lu (2015) | The study analyses the efficiency of banks in Taiwan by pointing out the marginal benefits of information technology (IT). In addition, considering the Basel III Accord, the impact of some proxies for risk on efficiency is measured. Most banks need to improve their returns to scale on IT inputs. The effect of risk proxies on efficiency was not universal in the study. | 0 |
| 38 | Nguyen et al. (2016) | The authors measure the cost efficiency of 32 Vietnamese banks in the period from 2000 to 2014, verifying the impact of two reforms in the banking sector, namely: partial acquisition by foreign banks and entry into the stock market. In addition, the study also analyses the impact of other environmental variables. Efficiency showed a slight upward trend in the period. Banks listed on the stock exchange or partially acquired by foreign capital presented better efficiency ratios. | 3 |
| 39 | Rayeni and Saljooghi (2016) | The authors investigate the relationship between efficiency and risk through a three-stage model in a case study with 14 branches. Risk causes banks to seek enhancement of their operations, thereby increasing their technical efficiency. Therefore, risk is positively related to efficiency. | 2 |
| 40 | Stewart et al. (2016) | The study analyses the determinants of efficiency of Vietnamese banks from 1999 to 2009. The largest banks are more efficient than the medium and small banks, with the latter being the most inefficient. Profitability had a positive impact on efficiency, while the number of branches and number of years in operation had the opposite effect. As far as global efficiency is concerned, private banks are more efficient than state owned banks. | 15 |
| 41 | Wanke et al. (2016) | The study estimates, through a two-stage model, the impact on virtual efficiency of M&A of Mozambique’s banks and also analyses the results taking into account whether the bank is state owned or has foreign control. The results indicate that control of the bank (state or foreign) affects efficiency and that mergers should occur between banks of different type of controls. M&A involving the analysed banks may lead, in most cases, to the situation of decreasing returns of scale. | 3 |
| 42 | Wanke et al. (2016) | The study uses a new Fuzzy-DEA model to evaluate bank efficiency in Mozambique for the years 2003–2011. Several aspects explain bank efficiency in Mozambique as, for instance, labour price, capital price and deposits. The effect of the environmental variables was ambiguous, depending on the degree of uncertainty of the model. Banks should reduce the number of employees and make initiatives to leverage capital. | 5 |
| 43 | Aggelopoulos and Georgopoulos (2017) | The authors analyse the efficiency of a bank’s branches in Greece during different periods of the economy, taking into account expansion followed by strong recessions The study also verifies how efficiency has behaved over the years. Banks’ efficiency deteriorated at the beginning of the recession, and especially as it deepened. | 3 |
| 44 | Alhassan and Tetteh (2017) | The study examines the impact of exogenous variables on the efficiency of 26 Ghanaian banks in the period 2003 to 2011. A high level of inefficiency among Ghanaian banks is evident, mainly due to pure technical inefficiency. The size of the bank positively influences efficiency only to a certain degree, due to economies of scale. Market concentration, leverage, and loan loss provisions are other significant factors identified as determinants of efficiency. | 1 |
| 45 | Azad et al. (2017) | The study evaluates and optimizes the productivity of employees from of the Bank of Industry and Mine in Iran by integrating DEA with AHP with quantitative and qualitative indicators. The results specify that the most inefficient branches are related to low work quality and high training hours. | 2 |
| 46 | Farandy et al. (2017) | The paper investigates the impact of exogenous variables on the efficiency of Islamic commercial banks in Indonesia from 2011 to 2014. The actual average efficiency of Islamic commercial banks in Indonesia is 91.82%. Assets and ROA had a positive impact, while the number of branches negatively affected the bank’s efficiency. | 1 |
| 47 | Fukuyama and Matousek (2017) | The authors extend the two-stage Network DEA (NDEA) by proposing a banking revenue function. In addition, the Nerlove model is also applied to identify bank inefficiencies. The results indicate that the Japanese regional banks did not reach the optimum point in their productive processes. The main cause of bank inefficiency is allocative efficiency. Capitalization and risk had a negative effect on efficiency. | 15 |
| 48 | Gulati and Kumar (2017) | The authors measure the operational and intermediation efficiency of 46 Indian banks through a two-stage Network DEA model, in addition to a bootstrapped truncated regression to verify the impact of variables on these indices. The overall efficiency of the sector needs improvement in the two stages analysed. Larger and private banks showed better results. | 2 |
| 49 | Kamarudin et al. (2017) | The paper analyses the determinants of productivity of Southern Asian Banks. National and foreign Islamic banks showed an improvement in Total Factor Productivity Change (TFPCH). Among the exogenous variables analysed, capitalization, liquidity and world financial crisis had a significant influence on the level of productivity level of banks. | 1 |
| 50 | Kong et al. (2017) | The authors propose an extension of the two-stage DEA model developed by Chen et al. (2004), making it possible to work with negative data and undesirable outputs. Operational efficiency, calculated in the first stage, is statistically smaller than profitability efficiency, measured in the second stage. | 1 |
| 51 | Shi et al. (2017) | The authors propose a model to estimate and decompose possible M&A gains from Chinese banks. The results show that banks can improve their operations, mainly in relation to technical efficiency, when engaging in M&A. In contrast, M&A have a negative impact on scale efficiency. | 3 |
| 52 | Wanke et al. (2017) | The authors analyse the virtual efficiency of M&A of South African banks. In addition, the impact of contextual variables on these efficiency indices is tested. M&A tend to be beneficial to banks, increasing technical efficiency, especially in terms of production. M&A gains are larger when both banks are local. | 6 |
| 53 | Chen et al. (2018) | The paper presents an innovative DEA model with SVM in the second stage in order to segregate efficiency groups. The study also analyses the effects of different context-related variables on efficiency indexes. For the sample of Chinese banks, efficiency is related to domestic origin and enlisting in the stock market. However, results show that performance of the Chinese banking sector is low. | 3 |
| 54 | Du et al. (2018) | The authors analyse the impact of earning asset diversification on Chinese bank efficiency from 2006 to 2011. In addition, they proposed an innovation on the method by extending the bootstrap model of Simar and Wilson (2007) Chinese banks could improve their efficiency with an increase in the diversification of their asset portfolios. | 1 |
| 55 | Fernandes et al. (2018) | The study measures the efficiency of banks in peripheral countries in the Eurozone and examines the effects of determinants of risk on bank performance over 2007–2014. Results indicate that higher levels of liquidity and credit risk negatively influence efficiency, while capital and profit risk have a positive impact on banks’ performance. The crisis tends to amplify the effect of bank risk. | 0 |
| 56 | Ouenniche and Carrales (2018) | The authors investigate the efficiency of 109 UK banks in the period 1987–2015, through DEA with a regression feedback mechanism. Several types of DEA model were used as well as different orientations. The proposed model increased the discriminatory power of the DEA. The SBM presented more consistent results than BCC and CCR. | 0 |
| 57 | Huang et al. (2018) | The study extends the NDEA to the Copula-Based Network SFA model, with application to US banks. The proposed model made it possible to overcome the convergence problem specifically when phenomena are subject to highly nonlinear simultaneous equations. The inefficiency of banks comes mainly from the first stage. | 1 |
| 58 | Xu (2018) | The study measures the efficiency of Chinese commercial banks and assesses the impact of foreign capital participation on efficiency. Banks with foreign capital tend to be more efficient, even if this share is owned by minority shareholders. In addition, efficiency is also influenced by macroeconomic factors. | 0 |
| 59 | Zhou et al. (2018) | The authors develop a dynamic two-stage DEA-SBM model to identify the sources of inefficiency of Ghanaian banks. Banks’ efficiency ratios were considerably low. The biggest source of inefficiency is in the first stage, called the productivity stage. | 0 |