Defining Influential Factors of Capital Adequacy Ratio: An Examination upon Turkish Banking Sector (2006/Q1-2019/Q1)

Capital adequacy ratio (CAR) of the Turkish Banking Sector (TBS) decreased dramatically from 30.9% in 2003 to 17.1% as of May 2019. This figure shows that although TBS has still a relatively high CAR compared to many countries, unfortunately there is a decreasing trend. A downward trend in CAR constitutes risks due to the limit of providing credit s. Therefore, the level of CAR has importance for making a positive contribution to sustainable economic growth. So, influential factors of CAR should be determined first. In this context, Multivariate Adaptive Regression Splines (MARS) method, 14 explanatory variables, and quarterly data are used for the period of 2006/Q1-2019/Q1. It is determined that credits/total assets ratio, legal equities, risk weighted assets, nonperforming loans (NPL), NPL/total credits ratio, and credit/deposit ratio are influential factors on CAR in Turkey.


Introduction
Countries have different financial systems from each other. As general, there are two types of financial systems, which are the bank-based and the market-based systems. Most of the countries, including emerging countries and Turkey, has a bank-based financial system which means that these countries provide most of the needs of funding for economic activities from banks (Kartal et al., 2018, Kartal, 2019. For this reason, banks and banking are important in such countries (Dinçer et al., 2016).
There is too much legislation regulating banking sectors and banks due to the fact that they are so much important for countries. That is why developments in financial systems and banking sectors have importance for the macroeconomics of countries (Kar et al., 2008). Similarly, Banking Law (BL) numbered 5411 regulates Turkish Banking Sector (TBS) strictly. Also, with the authorization of BL, Banking Regulation and Supervision Agency (BRSA) has made a variety of secondary regulation on TBS (Kartal & Çoban Çelikdemir, 2019). In addition to BRSA, also other regulatory bodies could make secondary regulation on TBS.
Although all regulation, unfortunately, there have been crises because of the fact that regulation and supervision system on banking sector does not function (Mishkin, 1999;Kılcı, 2017;Çam & Özer, 2018). Depending on the crisis like 2000 and 2001 crisis in Turkey and 2008 global crisis which started in the United States of America, national authorities and international institutions have intensified their efforts and published additional principles named as Basel III criteria which regulate the banking sector in a much more efficient manner. The most important issue in Basel III regulation is how to calculate adequate capital, i.e. CAR. Basel Committee on Banking Supervision would increase tier 1 capital/ risk weighted assets (RWA) ratio to 4.5% and tier 2 capital/RWA to 7% (BIS, 2010). This is a challenging process for some countries, because of increasing capital or decreasing RWA requirement. With the varying of transition period according to countries, Basel III process was initiated in 2013, regulations have been made in time, the transition period has been completed, and Basel III has gone into effect in 2019 fully in Turkey.
All countries have been in stress because of this hard transition period. When examining the development trend of CAR in Turkey, it can be seen that it has decreased from 30.9% in 2003 to 17.1% as of May 2019 (BRSA, 2019). As mentioned, CAR is important because it affects the capacity of providing credits of banking. When CAR is decreasing and reaching the legal limit which is 12% in Turkey, then any bank should decrease providing credits if the bank cannot increase their capital. For this reason, having a high CAR for both TBS and Turkish banks is crucial and definitely keeping the CAR high is very important.
Because of providing most of the financing source by banks in Turkey, decreasing trend of CAR has a risk. To be able to sustain financing support of banks to the economy, stopping decreasing in CAR is a requirement. On the other hand, increasing CAR is very important if much more credit providing by banks is desired. However, influential factors of CAR should be determined firstly. For this aim, MARS method, 14 explanatory variables, and quarterly data are used for the period of 2006/Q1-2019/Q1 to determine influential factors on CAR in Turkey. As far as it is known, MARS method is used first time in defining influential factors on CAR. So, it is thought that this study is a pioneer researc in terms of this characteristic.
This study consists of four parts. After the introduction, part 2 reviews the related literature upon capital adequacy ratio in Turkey and some other selected countries. Part 3 includes the data, methodology and research results. Part 4 summarizes the results.

II. Literature Review
There are studies in the literature handling CAR in Turkey and other countries. A variety of independent variables such as asset growth, asset quality, bank size, credits, credits/total assets, foreign exchange rates (FER), inflation, interest rates, leverage, liquidity, net interest margin (NIM), NPL, provisions, provisions/total credits, return on assets (ROA), return on equity (ROE), RWA and total assets are taken into consideration in these studies.
Some studies in Turkey examined the whole banking sector while some of the others analyzed some selected banks. Sayılgan & Yıldırım (2009)  In addition to studies examining whole TBS, some of the studies examined a part of the banks In other studies, Berger (1995) examined the USA for the period of 1983 and 1992 via granger causality test and concluded that CAR is positively related with ROE between 1983 and 1989, whereas it is negatively related with ROE between 1990 and 1992. When evaluating studies taking place in the literature and summarized above, it is defined that the effects of several independent variables on CAR are examined and these variables have either positive or negative effects. On the other hand, determinants of CAR are the focal point of a variety of different researchers. Furthermore, it is also identified that various methodologies were used in these studies such as ARDL Bounding Test, Fourier Approach, Granger Causality Test, GMM, Regression, Panel Data Regression, and Panel Data Analysis. This situation indicates that, a new method could be used to identify the determinants of CAR. With the help of this issue, it can be possible to contribute to the literature. So, MARS method is preferred in this study.

III. Defining Influential Factors of CAR in Turkey a. Data and Methodology
To determine which factors have an influence on CAR in Turkey, MARS method, 14 explanatory variables, and quarterly data are used for the period of 2006/Q1-2019/Q1. This period is selected, because data of credit/deposit ratio are not available before 2006, which is evaluated as an important independent variable in the analysis. Data regarding dependent and independent variables are gathered from BRSA (2019) and Central Bank of the Republic of Turkey (CBRT) (2019).

b. MARS Method
MARS Method is developed by Friedman in 1990s. MARS method is a non-parametric method and one of the machine-learning techniques. So, it does not include any restrictive assumptions (Friedman, 1991).
There are no assumptions among dependent and independent variables in the MARS method. In searching the effects of independent variables on dependent variables, MARS also uses interactions between variables, and the effects of these interactions on dependent variables (Goh et al., 2017;Liu, 2018).

MARS model is formulated below:
In equation (1), "Y" represents the dependent variable, whereas independent variables are shown as X. On the other side, 0 demonstrates the constant term and ( ) describes basis function. Therefore, represents the coefficient of n. basis functions (Friedman, 1991).
MARS method consists of two steps. In the first step, all possible models are produced by using independent variables until reaching maximum basis functions. In the second step, the best model is selected by eliminating some basis functions from the most complex model. The best model is the one that has the lowest Generalized Cross Validation (GCV) value and the highest GCV R 2 (Sephton, 2001).

Mustafa Tevfik Kartal
Emerging Markets Journal | P a g e |19

c. Independent Variables
In the literature, a variety of independent variables have been used to determine which factors affect CAR. Some of these variables are summarized in Table 1.  (2015) Taken into consideration data availability for independent variables, equity; RWA, credits, NPL, provisions, net profit, ROA, ROE, USDTL, interest rate, NPL/total credits, credits/total assets, RWA/total assets, are selected as independent variables. Also, with the thought that credit/deposit ratio has importance for TBS, it is included as an independent variable. Hence, a total of 14 variables are included in the study. Details of variables are included in Table 2.

d. Analysis and Empirical Results i. Descriptive Statistics
In this study, quarterly data for the period of 2006/Q1-2019/Q1 are used. So, the number of observations is 53 and descriptive statistics are included in Table 3.

ii. CAR Estimation Model Findings
In the first step of MARS analysis, all possible basis functions are produced by using 6 independent variables which affect CAR. In this process, 18 functions are produced totally, which includes the best complex function. Details of produced 18 functions are included in Annex 1.
In the second step of MARS analysis, 12th model is determined as the best one which has the lowest GCV value and the highest GCV R 2 value. Important splines between CAR and independent variables are included in Annex 2. In the best model, there are 14 basis functions using 6 independent variables and details of the best model are included in Table 4.

Mustafa Tevfik Kartal
Emerging Markets Journal | P a g e |21 As it can be seen from Table 4, the probability value of the F test is 0.000, which means that the model is statistically significant. On the other hand, the explanatory value (R 2 ) of the model is well above the acceptable limits with the value of 0.972.
As a result of the analysis, the importance level of independent variables in terms of explanation of CAR change in Turkey as included in Table 5. As it can be seen from Table 5, the most important variable in terms of CAR in Turkey is credits/total assets ratio. Other important variables are equity, RWA, NPL, NPL/total credits, and credits/deposit ratio respectively. On the other hand, other variables which are included in the analysis do not have an effect on CAR for the period of 2006/Q1-2019/Q1. First explanatory variable on the change of CAR is CRDSTA. The details of the basis functions regarding CRDSTA are included in Table 6.  Table 6 shows that, the variable takes place in 3 basis functions. BF17 has a negative coefficient (-0.119). When CRDSTA is above 60.06% and CDR is below 121.61%, CRDSTA has a negative effect on CAR. On the other, when CRDSTA is below 60.06%, CRDSTA does not affect CAR.
Another independent variable that affects CAR is EQITY. The details of the basis functions regarding EQITY are included in Table 7.  Table 7 shows that, the variable takes place in 5 basis functions. BF3 has a positive effect (coefficient: 0.072) if EQITY has a value above 129.88 billion TL. On the other side, BF4 provides information that, this effect becomes negative when EQITY is between 129.88 billion TL and 103.23 billion TL, in addition to CRDSTA being below 54.49%. Also, according to BF9, the effect of EQITY on CAR is positive because of the positive coefficient (0.032). However, if EQITY is under 103.23 billion TL and CRDSTA is above 54.49%, EQITY does not affect CAR.
Third independent variable that affects CAR is RWA. The details of the basis functions regarding RWA are included in Table 8.   Table 8 shows that, the variable takes place in 5 basis functions. BF12 has a positive effect (coefficient: 0.005) if EQITY has a value below 559.08 billion TL and CRDST is below 54.49%. On the other side, BF11 provides information that this effect becomes negative when RWA is between 559.08 billion TL and 259.53 billion TL. However, if RWA is under 259.53 billion TL, RWA does not affect CAR.
Fourth independent variable that affects CAR is NPL. The details of the basis functions regarding NPL are included in Table 9.  Table 9 shows that, the variable takes place only in 1 basis function. BF5 has a negative effect (coefficient: -0.360) if NPL has a value above 18.65 billion TL. However, if NPL is under 18.65 billion TL, NPL does not affect CAR.
Fifth independent variable that affects CAR is NPLTC. The details of the basis functions regarding NPLTC are included in Table 10.  Table 10 shows that, the variable takes place in 4 basis functions. BF14 has a negative effect (coefficient: -0.028) if NPLTC has a value below 3.42% and EQITY is above 129.88 billion TL. On the other hand, BF19 has a positive effect (coefficient: 0.022) if NPLTC has a value above 3.43% and RWA is below 259.53 billion TL.
Last independent variable that affects CAR is CDR. The details of the basis functions regarding CDR are included in Table 11.

IV. Conclusion
Determining influential factors of capital adequacy ratio has crucial importance for the security of the banking sector and financial system, and also for sustainable economic growth and financial stability. The main cause underlying these is that, banks are the main financing source in some countries which have a bankbased financial system structure. So, countries try to make the banking system stronger so that they can continue to finance economic activities. However, a downward trend in CAR produces risks due to the limit of providing much more credits of banks. Therefore, countries should determine influential factors on CAR firstly.
This study aimed at defining influential factors of CAR in Turkey. In this context, 14 independent variables are selected by benefitting from similar studies taking place in the literature. Also, quarterly data for the period of 2006/Q1-2019/Q1 are gathered and analyzed by MARS method.
As a result of the analysis, it is defined that credits/total assets ratio, legal equities, risk weighted assets, nonperforming loans (NPL), NPL/total credits ratio, and credit/deposit ratio have an effect on CAR in Turkey. According to the analysis, the most important factor is the credits/total assets ratio. CAR decreases when credits increase much faster than legal equities.

Mustafa Tevfik Kartal
Emerging Markets Journal | P a g e |23 Also, there is a similar relationship between legal equities and risk weighted assets.
Another important variable is NPL. It can be said that, CAR decreases if NPL exceeds 18.65 billion TL. However, the current NPL is around 106 billion TL in Turkey. Unfortunately, this condition causes decreases in CAR. Also, NPL/total credits ratio has negative effects on CAR. On the other hand, when it is below 121.61%, credit/deposit ratio has a negative effect on CAR by interacting with credits/total assets ratio.
With the evaluation of analysis results, it can be concluded that some negative developments in analyzed independent variables have been causing a negative effect on CAR. To prevent CAR from decreasing much more from the current level, which is 17.1% as of May 2019, negative effects should be prevented in the mentioned variables. For instance, an increase in NPL should be stopped or the sale of NPLs by banks should be eased. Hence, the negative effects of NPL on CAR could be prevented. Also, the negative effects of credit/deposit ratio on CAR could be prevented by easing the collection of deposits of banks. Also, it is an important point to be stated that the accumulation of legal equities should be sustained so that the balance between legal equities and risk weighted assets could be kept. Also, some new precautions could be developed by banks and regulatory authorities. Of course, those precautions should be deployed in time so that the Turkish banking sector could benefit from these. Hence, banks could provide more credits for financing and supporting economic growth in Turkey.
Besides this study, new studies such as examining why NPLs have been increasing so much in recent times in Turkey could be studied and it is thought that these studies could be beneficial in developing the literature. Also, new statistical and econometrical methods could be used in these forthcoming studies.