As elsewhere in the world, the banking sector is a vital part of the financial system in Turkey. In the Turkish economy, the banking sector’s total assets realized was USD 823 Billion, and the ratio of total assets to GDP was around 115 percent in 2020. In addition, the loan volume of the banking sector was USD 482 billion, and the percentage of loans to GDP was 74 percent in the same year. These figures underscore the significance of the banking sector in the Turkish economy.

In studies about the banking sector in developing countries, they have always been attempts about determining the effects of market structure, concentration, competition, financial liberalization, mergers and acquisitions and the internal factors on the efficiency figures. One of the first researches in developing countries is that by Bhattacharyya et al. (1997), study on the Indian banking sector. In that study, the production activities of 70 commercial banks were analyzed during the financial liberalization term. The results showed that state banks had the highest efficiency values (Bhattacharyya et al., 1997). Rezvanian and Mehdian (2002) studied efficiency analysis on the production performance and cost structure of the Singapore banking sector. They showed that there were scale economies in the sector and that the cost inefficiency was caused equally by allocative and technical inefficiency. Another study from India analyzed the effects of liberalization on the CPE in India. The study showed that the decrease in profit efficiency (PE) was due to allocation ineffectiveness. High cost efficiency (CE) and low-PE were indicators of income inefficiency in banking activities. According to the study, bank size, ownership, product variety and positive economic indicators were significant variables that influenced efficiency values (Das & Ghosh, 2009).

The study on banks in Latin America compared foreign and domestic banks in terms of the CPE. These studies, conducted in 16 countries and 427 banks, showed that environmental factors were at the forefront concerning efficiency differences between states, and the most critical ineffectiveness was income-related (Kasman et al., 2005). Another research approach analyzed liberalization effects on bank efficiency values in Southeast Asian countries. The results revealed a positive relationship between a high liquidity ratio and high-efficiency values and that the efficiency of public and family holding banks was lower than private ones (Williams & Nguyen, 2005). The effect of privatization and foreign ownership on efficiency values in Chinese banks has also been determined. It was found that foreign banks had the highest efficiency values in China (Berger et al., 2009). Kocisova (2014) analyzed the cost, income, and profit efficiencies of Czech and Slovak commercial banks between 2009- 2013 by using the DEA method. According to the results, the Czech and Slovak banks were more efficient in terms of income than cost and profit efficiencies. Sufian et al. (2016) found that domestic banks showed higher efficiency values ​​than foreign banks in their studies on the Malaysian banking sector. However, contrary to most of the findings in the literature, their study revealed that capital market movements negatively affected the technical efficiency of the banking system.

The first significant work on efficiency analysis in the Turkish banking sector was the work of Zaim in 1995. In the study, the technical and allocative efficiencies of banks between 1981 and 1990 were calculated using the DEA method, and the effects of liberalization on efficiency values were examined. The findings showed that the reforms influenced technical and allocative efficiency, and the state banks worked more effectively than private banks (Zaim, 1995). Following Zaim (1995), there were other studies by Yolalan, (1996), Yildirim (1999), Cevdet et al. (2007). Jackson and Fethi (2000), Işik and Hassan (2002), and Demir et al. (2005) on the efficiency of Turkish banks by focusing on financial liberalization, ownership, and scale efficiency. According to the results, the effect of liberalization on the sector was limited, and scale inefficiency existed in the sector (Yildirim, 1999). There was however, a positive relationship between bank size and the CPE (Jackson & Fethi, 2000). Competition in the credit and deposit market was below the optimal competition. The oligopolistic structure of the sector had decreased efficiency (Işik & Hassan, 2002).

Studies on the Turkish banking industry mainly focused on the cost side of efficiency; however, findings showed that profit efficiency remained limited. One of the few researchers on profit efficiency was the study by Isik and Hassan (2002). They found that Turkish banks were highly profit efficient, and the link between cost and profit efficiency was shallow. This revealed that high-profit efficiency did not require high-cost efficiency in the industry. Gunalp and Celik (2004) investigated the relationship between efficiency and competition in the Turkish banking industry between 1990 and 2000. By using the Stochastic Boundary approach, they found a positive correlation between efficiency values and profitability. Abbasoglu et al. (2007) also analyzed the efficiency values of the Turkish banking industry between 2001 and 2005, using the stochastic boundary approach. According to the study results, it was concluded that the concentration in the sector increased, the level of competition followed a fluctuating course, and the sector displayed a monopolistic competitive market structure. Fukuyama and Matousek (2011) analyzed the changes in the efficiency levels of the banking industry during the two crisis periods which occured in Turkey between 1991 and 2007. Unlike the other studies, the “two-stage network model” was used, and the efficiency values ​​were determined only by the VRS method. Yilmaz (2013) conducted a one stage efficiency analysis of the Turkish banking industry between 2007 and 2010 using the DEA method. According to the results, it was determined that domestic banks were more efficient than foreign banks and that the 2009 global crisis harmed efficiency scores. Gunes and Yildirim (2016) concluded that the Turkish banking industry, which has close relations with European countries, was not adversely affected by the 2010 European banking crisis and 2008 financial global crisis. Furthermore, it was observed that the cost efficiencies of Turkish banks did not decrease during these two crises. In the study by Batir et al. (2017), it was seen that based on the data collected between 2005-2013, participation banks had a higher efficiency value than traditional banks. Partovi and Matausek (2019) analyzed the efficiency values in the Turkish banking system through the effect of the NPL. This study showed results supporting the “bad management” hypothesis and revealed that the efficiency values of banks differ according to the capital structure. One of the most recent Turkish banking system studies has been the work by Ozbey and Akan (2021). The study which covered the 2000-2018 period determined that the most efficient banks were private banks, and the effect of personnel expenses on efficiency values was high (Ozbey & Akan, 2021).

This study makes important contributions to the literature in several ways. Firstly, it differs from previous researches in the Turkish banking industry by analyzing the profit efficiency, which has been rarely studied in recent years. While previous studies (Günalp & Çelik; 2004, Abbasoglu et al., 2007; Fukuyama & Matousek; 2011, Gunes & Yildirim; 2016, Batir et al., 2017; Partovi & Matausek, 2019; Ozbey & Akan; 2021) mainly focused on cost efficiency, this study calculates both cost and profit efficiency separately for commercial banks operating in the Turkish banking system. This comprehensive analysis provides insights into the structural and capital-based changes in the banking system and their impact on efficiency.

Secondly, the study uses both variable returns to scale (VRS) and constant returns to scale (CRS) methods to calculate efficiency values. As each method has its strengths, this approach helps to determine which method is more appropriate to calculate efficiency in the Turkish banking system. Thirdly, the study aims to determine the effect of bank ownership on efficiency values. As Turkey is considered one of the developing economies, it is crucial to understand the efficiency differences between private foreign banks and public banks in the country. The study by Ozbey and Akan (2021) filled a gap in the literature by providing recent data on this aspect. Fourthly, the study included the diversification effect of technological developments, which has been much neglected in previous studies (Günalp & Çelik, 2004; Abbasoglu et al., 2007; Gunes & Yildirim, 2016; Partovi & Matausek, 2019; Ozbey & Akan, 2021). The analysis included the types of activities that were becoming increasingly important in providing non-interest incomes in banking, and it had been able to determine the effect of diversification on profit efficiency. Finally, this research provides the most recent and longest-term (2006-2020) analysis of efficiency in the Turkish banking system. The study’s findings offer insights into the impact of recent technological developments, intense competition, consumer rights, and structural regulations on the efficiency of banks. The results could be useful for policymakers, regulators, and practitioners to understand the current state of the banking sector in Turkey and help them make informed decisions to improve its overall efficiency and stability.

The efficiency of a unit is obtained by comparing the realized inputs and outputs of the unit with the optimum inputs and outputs. Efficiency is divided into technical and allocative efficiency (Farrell, 1957). Technical efficiency refers to a unit’s capability to get the most significant number of outputs from inputs. However, allocative efficiency refers to the unit’s ability to use inputs optimally according to their prices. Efficiency measures developed over a period of time are primarily split into two parts, namely non-parametric and parametric methods. The most basic non-parametric method is the DEA. The analysis is widely used in the banking sector to compare the efficiency performances of numerous banks, or to estimate the efficiencies between particular bank units. DEA studies are based on Farell’s linear convex hull approach in estimating the efficiency line. It is developed and applied to multiple inputs and outputs (Charnes et al., 1978).

In all DEA models, due to the nature of the efficiency measurement, the transformation of inputs (X1, X2, ….XN) to outputs (Y1,Y2,….YN) is defined, and then the economic decision making units (DMUs) are ranked from the most efficient to the most inefficient. For this purpose, the efficiency value is applied to the entire data set through a virtual efficiency frontier. The virtual efficiency value is estimated by the ratio of weighted outputs to weighted inputs. When the efficiency frontier is estimated, all data points are folded in a convex hull. The efficiency of the DMU which is above the efficiency frontier is evaluated as efficient, and the one below is evaluated as inefficient. The essential feature that distinguishes the DEA from other methods is that it does not require a mathematical or statistical form of production. Moreover, in contrast to parametric methods, since no production function is predicted in the model, false results are also eliminated due to the incorrect estimation of the production function.

In the analysis carried out in the present study, the R program is used to measure the cost-efficiency. The program calculates the cost efficiency values with the “cost.opt” method in the Benchmarking package developed by Bogetoft and Otto (2011). The cost efficiencies of banks were calculated separately according to CRS and VRS methods and the results are as presented in Table 5 and Table 6. The CE values ​​according to ownership are as given in Table 7. Table 7 shows that state banks had better CE values than private and foreign banks. Table 5 shows that the banking system experienced a decrease in cost and profitability efficiency after 2009 (See Table 5).

The Benchmarking package was used in R program to calculate the PE by the DEA method. The VRS method was used to calculate the PE. PE values for the whole sector are as presented in Table 5. As can be seen in Table 8, the PE values of state banks were higher than those of private and foreign banks. In general, it is observed that the PE values of ​​ Private banks were better than that of foreign banks. However, it is observed that foreign banks have performed better in recent years (see Table 8).

The periods with negative profits were excluded from the data. As the majority share of some private domestic banks was sold to foreign banks during the research period, they were moved from the private banks’ group to the foreign banks’ group. In the present study, the group with the lowest profitability was identified as the group of foreign banks. It is seen that foreign banks have been working with shallow profit margins after 2009, and they had losses in specific periods. Banks with negative profitability were excluded from the average. Even with positive profitability values, they had a very low PE (see Table 8).

The left side is uncensored, and the right side is censored with “1”: In this limitation, all negative values of the PE were taken as dependent variables, while the efficiency values with “1” were not taken. Therefore, the total numbers of observations decreased from 1311 to 1276 due to the negative values being uncensored. The left side is censored with 0 and the right side with “1”. In this limitation, the PE values were only accepted when they were between 0 and 1. The total number of observations dropped to 1124.

The left side was censored with 0, and the right side with “1.1”. Banks with a negative PE were excluded from the observation. Since there was no bank with efficiency values above 1.1, the observation set consisted of 1159 observations. The left-hand was uncensored, and the right-hand side was censored with “1.1”. Here, the aim was to determine the regression result by taking the banks with a “1” total efficiency value into the observations. Since there was no bank with an efficiency value above 1.1, the observation set was analyzed as 1311 without censorship.

The total data observations covering 57 quarter periods between 2006 and 2020 was 1311. Tobit regression was performed in four different censors. When the data was censored with 0 and 1 in the PE analysis, it decreased to 1124. The censorship of negative efficiency values was widespread in profitability analysis. However, since banks could reach total efficiency value, the values in which the regression analysis was censored with 1.1 on the right side were used. According to the results, the independent variables of SIZE, DEP, TERMDEP, LIQUIDITY, ASSETGRW, NPL, DSTATE and DPUBLIC had significant effects on the PE values.

Tobit regression analysis was repeated to determine the relationship between cost efficiencies and independent variables. As in the PE analysis, the data was censored from the left and right, and two different regression results were obtained. Since the banks with one efficiency value were excluded from the observation in narrow regression analysis, the number of observations fell from 1311 to 1026. When the banks with one efficiency value were taken into the observation series, 1311 observations could be analyzed. According to the results, independent variables that affected the cost efficiencies were; SIZE, DEP, TERMDEP, LOAN, ASSETGRW, NPL, CRAR, RWA, DSTATE, DCRISIS and DPRIVATE.

The results show a positive and significant relationship between SIZE and the CPE. Large banks were seen as having the ability to adjust their output scales more optimally, thus increasing their profitability. Larger banks also have the opportunity to diversify their risks. This result supports the efficient structure hypothesis and corroborates the findings of Hauner (2005), Berger et al. (1993), Berger and Hannan (1998), Isik and Hassan (2002), Kasman (2002), and Ozkan and Gunay et al. (2013). Growth in total assets (ASSETGRW) has a positive effect on the CE as well as profit efficiency. The results are consistent with those in previous research, such as in Das and Ghosh (2009), Jackson and Fethi (2000), and Kasman (2002). Furthermore, consistent with the efficient structure hypothesis, the results reveal that the growth rate in a bank’s asset structure enables that bank to offer tools to increase profitability. The results also reveal a positive and statistically significant relationship between a bank’s deposit market share (DEP) and the PE. If it is the case that a market concentration leads to high prices and profitability, the coefficient is expected to be positive—accordingly, the more open the market, the greater the relationship between profitability and market share. According to the results, the share of deposits also has a positive and significant effect on the CE. The market share in deposits also provides banks with the advantage of reducing their deposit costs. These results were also obtained in the studies by Stiroh and Strahan (2003) and Das and Ghosh (2009). The high rate of time deposits in total deposits (TERMDEP) can lead to high costs in an environment where interest rates are falling. Therefore, the coefficient of this variable is expected to be negative. As is consistent with the literature, such as in Das and Ghosh (2009), the results show that the share of time deposits in total deposits negatively affects the PE and CE of the bank.

According to these results, there is a positive relationship between the bank’s liquidity and PE. The high rate of liquid assets is a sign of bad cash management and leads to low interest-income. It is expected that this variable will harm the PE. The difference between the results obtained from present study and the studies in developed countries can be explained by the fact that banks can even profit from liquid assets because of Turkey’s high cost of short-term funding. Therefore, the banks prefer to hold a certain amount of liquid assets for caution and make a profit from it. The NPL variable is expected to be negatively correlated with the PE and CE. According to the results, there is a significant relationship between the NPL variable and the PE and CE. The results are consistent with the bad management hypothesis introduced by Berger and De Young (1997). Recently Partovi and Matousek (2019) found similar results for the Turkish banking industry. The rise in the loan ratio (LOAN) indicates the high-risk structure of the bank’s statements and the higher market share in the loan market. According to the efficient structure hypothesis adduced by Berger et al. (2008) and Das and Ghosh (2009), banks with higher loan ratios are expected to have higher efficiencies. According to the results, there is a positive relationship between the loan ratio and CE at a high significance level. This result shows that banks with higher loans/assets ratios can better control the costs of banks.

Empirical studies which included Berger and Humphrey (1997) and Casu and Molyneux (2003). Belas et al. (2019), Isik and Hassan (2003), and Fukuyama and Matousek (2011) showed that well-capitalized banks had higher efficiencies. The results show a positive but shallow relationship between the CE and capital adequacy ratio (CRAR). It is expected that the banks with lower capital funding costs will increase due to the high leverage ratio. In addition, they are expected to be more vulnerable to possible credit risks. According to Berger and Deyoung’s (1997) bad management hypothesis, increasing the riskiness of a bank harms its profitability. The results show a negative relationship between the CE and the risk-weighted assets (RWA) ratio. In the analysis, DSTATE dummy was used for the control variable of bank ownership. The results show that state ownership affects the PE and CE negatively and significantly. This is because the results reflect the market regulator functions of state banks. This result is inconsistent with Zaim (1995) ‘s findings, but supports that of Isik and Hassan (2002b) which found that private banks were more efficient than state banks. This result is obtained because the state banks finance projects that private banks are unwilling to finance due to their market regulatory and social responsibility. The impact of the 2008-2009 mortgage crisis on Turkish banking started to be seen in 2010.

Therefore, 2010 is included in the model as a crisis dummy variable. According to censored Tobit regression results, positive relation is found between the Crisis dummy and cost efficiencies. This result is not compatible with the research carried out in Ozkan-Gunay et al. (2013) and Aysan and Darendeli (2010), Yilmaz (2013), Gunes and Yildirim (2016) which showed that the global crisis harmed the Turkish economy. This result shows that the regulations after 2002 in the banking system made the banking industry more resistant to crises. The DPRIVATE dummy variable is used to see the effect of private banks on efficiency values. According to the Tobit regression result, private banks harm cost-efficiencies with an 8 percent coefficient. Therefore, the market discipline hypothesis says that private banks have higher efficiencies. However, the results do not support the hypothesis and reveal the opposite results as in Isik and Hassan (2002). Therefore, it can be concluded that intensive banking system regulations negatively affect private banks’ cost efficiency.

The DPUBLIC is the Dummy variable that measures the effect of being listed on the stock market on the PE and CE. The year in which the bank offers its shares to the public takes the value 1. Publicly traded companies are expected to have higher efficiency. The market discipline hypothesis tested by Berger and Mester (1997) and Das and Ghosh (2009) found that publicly traded banks had higher cost and profit efficiencies. However, Isik and Hassan (2002) found that public offering had a positive effect just on cost efficiency. Similarly, this study has found that if a bank was publicly traded, that bank would have a higher profit efficiency.

In this study, the CPE values of 23 commercial and deposit banks operating in the Turkish banking industry were analyzed by the DEA method. Quarterly data between 2006 and 2020 from these banks were collected and analyzed. In addition, Tobit regression analysis was also performed to identify the determinants of the efficiency values obtained. The study aimed to obtain the most recent and long-term results on profit and cost efficiency in the Turkish banking industry.

The results show that Turkish banks work with relatively higher cost efficiency than profit efficiency. In 2020, it was seen that the COVID-19 pandemic harmed the cost efficiency and profit efficiency of the banks studied. On the cost efficiency side, state banks were the category of banks least affected by the pandemic, while a 10 percent decrease in efficiency was observed in private banks.

On the other hand, in the first year of the pandemic, there was an average of 25 percent depreciation in private and foreign banks in terms of profit efficiency. In contrast, in the same period an increase was observed in state banks. The study found that during the pandemic period, state banks gained market share on the credit and deposit side and acted as a locomotive in the banking sector. According to multivariate the Tobit regression analysis carried out in this study, Total Assets, Deposit Share, Time Deposits, Liquidity, Asset Growth, NPL, Ownership and Publicly when traded significantly affect profit efficiency. Therefore, the results show that there are some particular implications. One of them is that the ratio of liquid assets to total assets was positively correlated with the efficiency values. This was in contrast to the findings in previous studies. The difference can be explained by the fact that the banks in the study could even get interest income from liquid assets because the short-term funding costs were relatively high in Turkey.

The second implication is that foreign banks need to perform better in both CPE. This result is in contrast to the findings in Berger et al. (2009) and Catalbas and Atan (2005). As a result of the decrease in PE, it is observed that the interest of foreign financial institutions in the Turkish banking sector in 2011, and in the intense acquisitions and mergers seen in 2002-2007, had decreased significantly. Consequently, some foreign banks have closed down their SME segments because of the high-risk and low profit. In addition, M&A can be seen in the sector because the size and deposit share have both affected PE positively.

Finally, the downward trend in the PE values in the Turkish banking sector is a warning signal on its financial health and stability. Therefore, to maintain the healthy growth of the banking sector and ensure the continuation of foreign capital inflows to the sector, it is essential to analyze, especially the factors affecting PE. Making new researches on the variables that are determinants in the CPE; detailing the research based on scales, segments and regions will give the parties concerned a broader view and enable a comprehensive analysis of the relevant issues involved.