PANEL STOCHASTIC FRONTIER ANALYSIS OF PROFITABILITY AND EFFICIENCY OF TURKISH BANKING SECTOR IN THE POST CRISIS ERA

This paper examines the efficiency and its relation to profitability in Turkish banking sector by employing Panel Stochastic Frontier Approach. In the post crises period, extensive structural changes have taken place and a great number of new developments have occurred, affecting the efficiency of banking sector. This is the first study that employs panel stochastic frontier approach for banking efficiency in Turkey. In this research, both cost and profit efficiency measures are estimated for the panel data consisting of 32 banks between 2002–2007. Results suggest that there is cost efficiency gain and convergence in the efficiency levels of banks. As another interesting result, foreign banks are less efficient and state banks are more efficient. This paper also analyzes the relation between efficiency and profitability and finds no robust relation between them. However, the bank size matters more for profitability.


Introduction
November 2000 and Februar y 2001 crises adversely affected Turkish economy and particularly Turkish banking sector. In the post-crisis period, extensive structural changes have taken place in Turkish banking sector. Interest of foreign banks for the Turkish market increased. Some new foreign banks entered into Turkish banking sector through acquisition, while existing foreign banks increased their operations. Foreign banks are expected to bring new practices and advance technology to the market and enhance competitive pressure in banking. Throughout 1990s Turkey experienced very high interest rates and accumulated huge debt stock surpassing Gross National Product. Consequently banks did not perceive any need to operate more efficiently given that they could earn enormous returns through financing government. In the post-crisis period, inflation, interest rates and debt stock started to decline. Eventually banks felt the need to rely more on essential banking activities to make more profit. Hence they had to operate more efficiently. As a result of these changes Turkish banks experienced profound transformations in their cost and profit efficiencies. These developments in cost and profit efficiencies shall have implications for the profitability of banks.
In this paper, we investigate the cost and profit efficiencies of Turkish banking sector in the post-crisis era by employing panel stochastic frontier approach. Our data set spans 2002-2007 period just before the global crisis. We further divided this period into 2 sub-periods as 2002-2005 (period of recovery and merger activities) and [2005][2006][2007] (period of growth and acquisition by foreign banks). According to our knowledge there are studies that employ stochastic frontier approach, but this is the first study that employs panel stochastic frontier approach to analyze the efficiency of Turkish banking sector for this period. Panel data has various advantages which significantly improve efficiency analysis compared to previous studies. Moreover we explore the relation between efficiency, size and profitability. Finally state banks are quite dominant in the banking industry in Turkey. Therefore we conduct the same analysis by excluding the state banks to implement sensitivity analysis.
In this paper we address the following questions: How does efficiency change over time? Is there a substantial efficiency improvement? Does the foreign banks prefer to buy more efficient banks? Is there efficiency gain in the banks acquired by the foreign banks? Is there a relation between profitability, efficiency and size?
The rest of the paper progresses as follows: Section 2 reviews the efficiency literature in banking sector and provide an overview of Turkish case in the post crisis period. Section 3 defines the data and explains the methodology and advantages of our model. Section 4 discusses the empirical results of efficiency and its relationship with size and profitability. Lastly section 5 concludes.

Literature review and Turkish case in retrospect
During the crisis period in Turkey, the banks which did not employ risk management techniques effectively had maturity and currency mismatch problems in their assets and liabilities. As a result of crises, interest rates increased sharply and Turkish currency rapidly lost value against other currencies. Hike in the interest rates especially hits the banks that had maturity mismatch problem in their portfolios. As a result of the increase in the interest rates, the assets of these banks also rapidly lost value and maturity mismatch in their portfolios did not allow the value of their liabilities to decrease by the same amount. Interest rates in domestic currency was higher than the interest rates in foreign currency. Therefore most of the banks had short position in foreign currencies and long position in Turkish currency. In fact, before the crisis most of the Turkish banks had both cur-rency and maturity mismatch problems in their portfolios. Furthermore contraction in economic activities engendered the rise in bad debt of banks.
In the aftermath of the devastating crises, Turkish banks were in a very uneasy situation. They made huge losses and some of them were on the edge of bankruptcy. As in many other developing countries, in Turkey banks are main financial intermediaries which channel saving into investment. This gives banks a major role in the capital accumulation and growth. Turkey urgently needed less fragile financial sector for consistent growth and economic prosperity. Hence Turkey initiated Banking Sector Reconstruction Program on May 15, 2001 to establish a competitive and healthy banking sector (see Al, Aysan 2006). In the scope of Banking Sector Reconstruction Program, capital structure of banks were strengthened, merger and acquisition activities encouraged. Furthermore Treasury helped the banks to close their short positions in foreign currencies while regulation and legislation were improved.  Table 1).  Table 2).  Table 3). There is a growing literature that investigates possible effects of foreign entry into the banking sector. Bonin et al. (2005) and Levine (2001) suggest that foreign banks increase efficiency of the banks by improving corporate governance. Moreover domestic banks acquired by foreign banks are upgraded by international rating agencies (Cardenas et al. 2003). Usually foreign banks bring new financial products and services, which enhance competition. Berger et al. (2000) show different results in the case of developed and developing countries about efficiency of foreign banks. Results suggest that foreign banks are more efficient in terms of cost and profit in developing countries and less efficient in developed countries compared to the domestic banks. Aysan and Ceyhan (2007) investigate the reasons for foreign bank entry in the light of push and pull factors. They suggest that Turkey's location (intersection of Europe and Middle East) increasing population and per capita income and EU accession process are the factors attracting foreign banks to invest in Turkey 2 . This literature reveals that foreign bank entry has effects on bank efficiency and structure. Hence it is quite interesting to analyze the period after the acceleration of foreign bank entry into Turkey.
There is also considerable literature on the relation between efficiency and profitability 3 . Turati (2003) analyzed this relation by examining correlation coefficient which he computed between efficiency scores and profitability. Abbasoglu et al. (2007) explore efficiency of Turkish banking sector and its relation with profitability. They found no robust relation between efficiency and profitability. There are also some studies that compares the efficiency of domestic and foreign banks. For example Isik and Hassan (2002a) analyzed efficiency of Turkish banking sector by Data Envelopment Analysis (DEA). They found that foreign banks are not more efficient than domestic banks.

Data and definitions of variables
We use the quarterly panel data of the all commercial banks of Turkey for the period 2002Q4-2007Q2. The data are taken from the Banks Association of Turkey (BAT).
There are 32 banks of which 3 are state banks, 13 are domestic banks, and 16 are foreign banks. We use two distinct dependent and seven independent variables consisting of four outputs and three inputs. Dependent variables are total cost (tc) and profit (p), or net income; and independent variables consist of outputs which are short term commercial loans (y1), long term commercial loans (y2), off balance sheet items (y3), and other earning assets (y4); and of inputs which are price of labor (p1), price of capital (p2), and price of funds (p3). Price of labor is the total expenditures on personnel and services, price of capital is total expenditures on physical capital divided by the book value of fixed assets and price of funds is total interest expenses divided by total funds borrowed. These variables are commonly used in the cost and profit efficiency of the banking sector literature 4 . Hence, we choose these variables in our study. As a measure of profitability we use two different measures return on asset and return on equity (see Table A.4 and Table A.5). Book value is taken as a measure of size.

Measure of efficiency, profitability, size and methodology
We can calculate efficiency by using either non-parametric (originating from operations research) or parametric approaches (econometric approaches). In a nonparametric (nonstochastic) approach like Data Envelopment Analysis efficiency is calculated by linear mathematical programming techniques. Parametric (stochastic) efficiency is calculated via a cost or profit function in which variables are costs or profits determined by input prices, quantities of variable outputs, random error and inefficiency. In our study, we use parametric approaches because of its two main advantages. In parametric approaches inefficiency is separated from statistical noise and we can use standard statistical tests on the variables to test their significance (Farsi, Filippini 2004). On the other hand non-parametric approaches do not allow this kind of statistical inference (Isik, Hassan 2002b).
In this study we calculate cost and profit efficiency following the paper of Isik and Hassan (2002b). Cost inefficiency is caused by using sub-optimal input combinations on a given output level while profit efficiency stems from using sub-optimal output level given the input and output prices. In other words cost efficiency shows how far a bank's cost is away from the bank that shows best performance if they produce under same conditions and same level of output. Profit efficiency shows how much bank is close to the highest amount of profit for its given level of output.
In this research we estimate both cost and profit frontier by time invariant panel stochastic frontier approach. We discuss the benefits of this approach over regular stochastic frontier models after introducing the model as follows:

________
In these equations tc it stands for total cost, π it stands for profit, γ fit stands for output, ρ kit stands for input, i indicates the bank, t indicates the time, l indicates the output, k indicates the input and υ it is a classical error term that follows a symmetric normal distribution. It is assumed that µ ji follows truncated half normal distribution and υ jit is independent of µ i , for j = 1, 2. Translog specification is employed in modeling both cost and profit function. In the empirical literature on bank efficiency translog specification is widely used. This functional form has various advantages, one of them is its flexible form which allows us to use Cobb-Douglas specification. Resulting 4 output, 3 input models for a given i th firm are as follows. Lastly we look at the advantages of panel data over cross sectional data in efficiency estimation. Schmidt and Sickles (1984) discuss the main advantages of panel data. Firstly there is no need to impose distributional specification on the efficiency term for consistent estimations. Secondly one can relax the assumption that inefficiency and input levels are independent. Moreover technical efficiency can not be consistently estimated in a single cross section, because its results heavily rely on distributional assumption on inefficiency 5 .

Cost and profit efficiency change
There are two important observations. There is a clear cost efficiency gain in Turkish banking sector in this period. Mean of efficiency scores increased from 0.74 to 0.91. One can also observe convergence in terms of cost efficiency. Standard deviation of cost efficiency scores declines from 0.06 to 0.04 (see Table 4 and Table A.1).
Apparently, profit efficiency roughly declines. However this can be attributed to the increased standard deviation between these two periods. Hence, in terms of profit efficiency divergence instead of convergence is the pattern. Recent developments in Turkey increased competition in the banking sector. Competition lowers the excess profits, which can affect profit efficiency.
When we look at the cost efficiencies of the banks for the period 2002-2005, the most cost efficient bank is AKB, whereas the least one is HBB. Among the highest ten cost efficient banks, all three of the state bank are included, only one of them is foreign banks, and the remaining six are domestic banks. Beside the banks that have the worst cost efficiency appear as follows: three are domestic banks, and the remaining seven are foreign banks. Hence, the overall cost efficiency of foreign banks is poorer in the period 2002-2005, whereas the state banks and domestic banks have better cost efficiencies. ________ Looking at the period 2005-2007, the highest ten cost efficient banks consist of two state banks, five domestic banks and three foreign banks. Lowest cost efficient banks that have least cost efficiency consist of two domestic banks, and eight foreign banks. In this period, again, foreign banks did worse in terms of cost efficiency, but they are better than their rankings in the former period. The efficiency of state banks remains almost same given that efficiency of TCZB declines whereas efficiency of THB increases almost the same amount. Furthermore the ranking of the overall cost efficiency of the domestic banks converges to median since the share of the domestic banks in least ten and highest ten declines.
For the overall cost efficiency ranking, in the 2002-2007 period, the state banks and the domestic banks are the most efficient, and foreign banks did the worst. All state banks are among the highest ten cost efficient banks, whereas nine of the least ten cost efficient banks are foreign banks. Beside domestic banks are almost above the median. Profit efficiency rankings of the groups of the banks are more homogeneous than the cost efficiency ranking. In the first period, 2002-2005, foreign banks were dominant among the highest ten profit efficient banks: five foreign banks, four domestic banks and only one state bank. On the other hand, in the second period the domination of the foreign banks is more apparent: eight foreign banks, and two domestic banks. State banks did worse in terms of profit efficiency compared to their cost efficiency ranking. In both periods their rankings are about the median.
Our results do not indicate any evidence supporting the idea that international investors look for higher efficiency in their acquisition decisions.
There are examples of banks that are inefficient but acquired. Banks that are acquired by foreign banks experienced efficiency increase. However in this period overall efficiency score of banking industry increases as well. In retrospect some of foreign banks experienced efficiency decline relative to other banks suggesting no clear-cut evidence in favor of efficiency improvement for the banks acquired by foreign banks (see Table 5). Lastly according to our results profit efficiency and cost efficiency are not related. Cost efficient bank can be profit inefficient and profit efficient bank can be cost inefficient. However we observe in general that in the first and second period profit and cost efficiency are negatively related (see Table 5, Table 6 and Table A.3).

Efficiency, size and profitability
We use book value of banks as measure of size and return on asset (ROA) and return on equity (ROE) as measures of profitability. We run fixed effect regression with panel data of 64 observations to examine the relationship between efficiency and profitability 6 . Our results do not suggest that there is a significant relation between cost efficiency, profit efficiency measures and profitability. We find however significant relationship between size and return on equity and return on asset suggesting that the size matters more for profitability in Turkey (see Table 7, Table 8 and Table A.2). We also run random effect regression with the same panel data and add dummy for foreign banks and state banks. Generally these dummies are insignificant while other results are very similar. Goodness of fit of our regressions are quite good considering that most of actual observations are in the confidence interval of our regression fit (see Figs. 1-4).

Sensitivity analysis
Turkish banking sector is known as state dominated sector. Although there are few state banks, their size is large. We conduct the same analysis to see the sensitivity of our results to the state owned banks. The findings show that our results are insensitive to the exclusion of state banks. Correlation between efficiency scores from the results of the analysis with the state banks and without the state banks are all more than 99 percent (see Table 9, Table A.6, Table A.7 and Table A.8). Furthermore we do not find a significant relation between efficiency and profitability which confirms our earlier results.

Conclusions
In this paper, we analyze cost and profit efficiency of Turkish banking sector in the post crisis era (2002)(2003)(2004)(2005)(2006)(2007)  The results of our study reveal that there is an increase in the cost efficiency in addition to convergence in the cost efficiency of banks. This finding shows that banks in Turkish market easily adopt new practices which enhance efficiency. When one bank discovers ways to increase its efficiency or a new more efficient bank enters into Turkish market other banks quickly imitate better technology. We also find that foreign banks including new entrants are less efficient. Our results also show that state banks are more efficient.
The results about state banks and foreign banks are quite interesting for the literature while they are in congruent with prior studies in Turkish banking sector.
We can not necessarily claim that banks acquired by foreign banks are more efficient banks. In the sample of banks acquired by foreign banks, there are efficient and inefficient banks. Efficiency of the banks acquired by foreigners increased. However there is an overall efficiency increase in this period anyway suggesting that these banks have relatively not performed better. We also analyze the relation between cost-profit efficiency, size and profitability by both fixed effect and random effect regressions. According to our results there is no significant relation between efficiency and profitability. However there is a positive relationship between efficiency and size. However we find significant relationship between size and profitability. Lastly we examine the sensitivity of our results for the exclusion of state owned banks. We conduct the same analysis by excluding the state owned banks. The findings confirm that our results are not sensitive to the exclusion of state owned banks.