The Effects Of Ownership Structures And Specific Characteristics On The Capital Structures Of Idx-Listed Banks

The present study observes the effects of ownership structures and specific characteristics on the capital structures of banks listed in Indonesia Stock Exchange (IDX). The author finds that the ownership structures and bank-specific characteristics (e.g., profitability, size and credit risk) do not have significant partial effect on the capital structures, while expense management does. This is consistent with a strand of previous studies including Haruman ( 2008), Yuke and Hadri ( 2005), Rista and Bambang ( 2011), Siringoringo ( 2012) and (Imas, et al., 2015).


INTRODUCTION
lays out banking-control perspective to classify bank ownership in Indonesia, which includes concentrated ownership, government ownership, private domestic ownership and foreign ownership. Large individual ownership indicates that bank ownership in Indonesia is concentrated into a number of owners. Managers, as a result, are simply subordinates to controller stakeholders. The major differences between government-controlled banks, domestic banks, mixed banks and foreign banks lie in the capitals and legal forms (Siringoringo, 2012). Capital structure policy deals with an optimal combination of using different sources of funds to finance an investment and a firm's overall operations to meet substantial financial goals and, in turn, to gain significant profits and values from the emerging market (Gitman, 2009). In addition to bank ownership, bank-specific characteristics, or internal factors, are taken into consideration to examine the capital structure decisions of the banks listed on IDX. Prominent bodies of literature have shed light on the relationship among the three domains with a wide variety of research findings. In the present study, the author breaks down the variables of bank-specific characteristics into profitability, size, credit risk and management expense. Jensen & Meckling (1976) establish the relationship between ownership structures and capital structures in terms of the percentage of share ownership by insiders (stakeholders) and outsiders (shareholders), in addition to debt and equity, when it comes to the most important factors in capital structures. As insider ownership rises, managerial ownership serves to align the interests of outside shareholders and managers (who act as agents as well as principals), and agency cost will decline. Hence, agency cost will rise with a reduction in managerial ownership. Bathala, et al. (1994) find the comparison regarding capital structure policy that the higher the insiders' proportions, the greater the desires to reduce the firm's cost of capital. In other words, firms no longer need to take on debts because the agency costs of debt will reduce as insiders own more shares. Furthermore, in the presence of large-block shareholders who buy stocks with a large amount of institutional ownership, firms exercise more control and external monitoring, which in turn leads to agency cost-reducing mechanism. One might expect that debt policy and institutional ownership can be a substitution-monitoring effect mechanism. Bathala, et al. (1994) explore this effect and find that institutional investors can encourage more effective monitoring services and mitigate the extent of opportunistic behaviors by managers. The monitoring of managerial activities may help reduce agency conflicts and become a substitute for debts. Banking institutions involve a very complex set of agency relationship. Common examples of this relationship include principals (shareholders) and agents (management), bank and debtors, and bank and regulators (Taswan, 2010). Myers (1984) suggests that funding needs are based on a certain hierarchy in choosing how a firm funds its growth projects as to minimize the likelihood of information asymmetry. Basically, a firm will prioritize internal, when available, over external financing. When external financing is required, a firm will prefer debt to equity owing to lower information costs resulting from debt issues. A firm will finally finance itself by issuing new equity shares as a last resort.

Asymmetric Information Theory
1. Cost of funds is minimized to the least possible level by setting up a certain composition. 2. Funds with low volatility and high stability are the bedrock of liquidity management. 3. The composition of funding sources holds the implementation of credit commitments and placement of other productive assets to the largest extent possible. Hypothesis Following the arguments above, the author tests potential explanation for the effect of ownership structures and bank-specific characteristics on capital structures through the following hypotheses: Hypothesis 1: The ownership structures of banks significantly affect their capital structure. Hypothesis 2: The bank-specific characteristics significantly affect their capital structure.

RESEARCH METHOD Operational Variables
The following table presents the variables observed in the present study: Ratio Secondary

Data Sources and Types
The present study relies on quantitative research, which includes secondary data. In secondary data analysis, the author analyzes the pre-existing data available from other sources and may have been used and published in previous researches, e.g., the balance sheets and income statements of publicly-traded banks from 2013 to 2016 and the proportions of bank stock ownerships. The population empirically selected for the study is the entire conventional commercial banks listed on IDX. Purposive sampling is applied based on the purpose of the study and the following characteristics of the population: 1. The conventional commercial banks operate from 2013 to 2016 and present their financial statements during the same period. 2. The banks have periodically released and consolidated the fully audited financial statements during the period.

Data Collection
The author gathers data to inform the research questions using library research, a disciplinary resource featuring a diverse array of scholarly journals, publications and the likes across the areas in need of investigation that highlight the subject of interest. = = = SAINS: Jurnal Manajemen dan Bisnis p- ISSN: 1978-2241e-ISSN:2541-1047 Maesaroh

Data Analysis and Hypothesis Test a. Data Analysis
In statistical modeling, multiple linear regression analysis enables the author to assess the effect of more than one predictor variable on a criterion or response variable. The simple form of regression equation to assess the association between these two types of variables is defined by: The calculation and interpretation of the correlation coefficient between the independent variables, e.g., ownership structures and bank-specific characteristics (rx 1 x 5 , rx 1 x 6, rx 1 x 7, rx 1 x 8, rx 2 x 5 ,rx 2 x 6,...... rx 4 x 8 ), are given by the following formula: b. Hypothesis Test The given model or equation considers a set of statistical inferences, both partially (individually) and simultaneously, across the variables with the testing criteria at a significance level of 5% (α =0,05).

Overall Test (F-Test)
The statistic outcome of the simultaneous association between independent and dependent variables is defined by the following hypothesis formula: H 0 :  xy 1 =  xy 2 = ....=  xy 8 = 0 H a : at least one  yx i  0 ; i = 1, 2, …, k The F-test for the overall or simultaneous significance is as follows: (2006) The above statistical test follows Snedecor's F-distribution with degrees of freedom v 1 = k and v 2 = n-k-1, where k = the number of independent variables. The F-test recognizes the following criteria:  If F cal ≥ F table with α=0,05, H 0 is rejected-there is a simultaneous effect of the independent variables on the dependent variable.  If F cal ≤ F table with α=0,05, H 0 is accepted-there is no simultaneous effect of the independent variables on the dependent variable.

Individual Test (T-Test)
When the test of the overall hypothesis formulation rejects the null hypothesis (H 0 ), at least one path coefficient is not equal to zero (p YXi ≠ 0). Under such circumstance, partial (individual) test is required to determine whether each independent variable, individually, is enough to create a significant relationship with the dependent variable. The hypothesis formula where partial path coefficient is assumed is defined by:  Gujarati (2006) Partial test between X variables (X 1 -X 8 ) and Y variable includes a two-tailed significance, given that the present study does not hypothesize a specific directional correlation (i.e., positive and negative correlation) between the two variables of interest. The criteria are as follows: If -t table >│t cal │>t table (α, n-k-l), H 0 is rejected-there is a significant partial effect of independent variable (X 1 ) on dependent variable (Y). If -t table <│t cal │<t table (α, n-k-l), H 0 is accepted-there is no significant partial effect of independent variable (X 1 ) on dependent variable (Y).

RESULT AND DISCUSSION Multiple Linear Regression Analysis
Prior to the procedure of regression model, we need to tap into a set of assumptions regarding linear regression that, in Gauss-Markov's term, fits into BLUE (Best Linear Unbiased Estimator) (Gujarati, 2011). a. Essentially, -best‖ is defined in a sense that regression line is the -best guess‖ at using a set of data to make a prediction. Regression line is necessary to express the pattern of relationship that relates two or more series of data. A line that fits the data well will be the one that minimizes the sum of errors. An error results from the observed value of a response variable that differs from the value predicted by the regression line. An efficient estimator, in addition to the -best‖ property, is unbiased. b. Statistical inferences in linear regression focus on β with the assumption that the relationship between the predictor X and the response Y is linear.
On average, is a linear estimator that expresses linear function that fits a predictive model to an observed data set of X values. OLS (Ordinary Least Square) estimates also minimize the squared residuals, thus creating linear estimates. c. An estimator is said to be unbiased if the estimator's expected value of β is not different from the true paramater value of β (β = β). Four principal assumptions, including normality test, multicollinearity test, heteroscedasticity test and autocorrelation test, are used to justify the linear regression models for the purpose of inferences or predictions. 4.1.1.1 The Effect of Government Ownership (X 1 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (Non Performing Loan) (X 7 ) and Management Expense (X 8 ) on Debt to Equity Ratio (Y) a. Normality Test The following figure shows a graphical method to decide whether the data come from a normal distribution. The graphical assessment of normality above shows the points track closely to the diagonal line, indicating that the data set is well modeled by a normal distribution.

b. Heteroscedasticity Test
The following figure also uses a graph for the examination of heteroscedascity. Figure 4.

Scatterplot of Heteroscedasticity
The scatterplot graph presented in Figure 4.2 shows that there is no obvious patterns in distribution, and the plots spread above and below zero on Y axis, thus indicating the absence of heteroscedascity. In other words, the regression model conforms to the assumption of homoscedascity.

c. Multicollinearity Test
To indicate the extent to which multicollinearity is present, VIF is calculated for each predictor using SPSS with the following output: The VIF value for each predictor is seen to be far less than 10, i.e., X 1 = 1,662, X 5 = 1,509, X 6 = 8,227, X 7 = 4,521, and X 8 = 4,570. Thus, multicollinearity is not present as two or more predictors included in the model are not significantly correlated so that the value of one cannot linearly be predicted by that of the other with a substantial degree of accuracy.

d. Autocorrelation Test
Linear regression model is tested for autocorrelation. The resulting statistical value d = 1,117 in SPSS (14.0 for Windows).

e. Analysis of Multiple Linear Regression Equation
After examining that the model assumptions are not violated, multiple linear regression analysis is run to measure the effect of Government Ownership (X 1 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (X 7 ) and Management Expense (X 8 ) on DER (Y). The aim is to identify the relationship among the variables and use this relationship to make predictions about the dependent variable based on the observed values of the independent variable in a causal inference. The multiple regression model is: Y =  + b 1 X 1 + b 5 X 5 + + b 6 X 6 + b 7 X 7 + b 8 X 8 + e Where Y = DER X 1 = Government Ownership X 5 = Profitability X 6 = Size X 7 = Credit Risk (NPL) X 8 = Management Expense  = Constant/ Intercept b 1,5,6,7,8 = Regression Coefficient e = Residual Variable Based on SPSS, the calculation of multiple linear regression yields the following output: Following the above output, the resulting constant and regression coefficient can be used in multiple linear regression analysis to build a regression equation: Y = 31,125 + 1,204 X 1 + 9,210 X 5 -1,564 X 6 -0,349 X 7 + 55,740 X 8 The interpretations are as follows:  = 31,125 If Government Ownership (X 1 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) take on zero ( If Size (X 6 ) increases one unit and the others are held constant, DER (Y) will decrease by 1,564 units. b 7 = -0,349 If Credit Risk (NPL) (X 7 ) increases by one unit and the others are held constant, DER (Y) will decrease 0,349 unit. b 8 = 55,740 If Management Expense (X 8 ) increases by one unit and the others are held constant, DER (Y) will increase 55,740 units.

g. Simultaneous Hypothesis Test (F-Test)
Below is a set of simultaneously-tested hypotheses: There is no significant effect of Government Ownership (X 1 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) on DER (Y) at the same time.
There is no significant effect of Government Ownership (X 1 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) on DER (Y) at the same time. Significance level α = 5%. The statistical test is F-test. The F statistical value using SPSS is presented below: In the Anova output, the author uses the F cal , which is 15,794, and compares it to the probability distribution of F-value. For α=5%, db 1 (degree of freedom) = k = 5, and db 2 = nk -1 = 28 -5 -1 = 23, the resulting F  The F cal , as it appears in the curve, is greater than the F table (15,794 > 2,640). H 0 is therefore accepted, indicating that the group of X variables (Government Ownership, Profitability, Size, Credit Risk, and Management Expense) is jointly significant in DER (Y).

h. Partial Hypothesis Test (T-Test)
A T-test, unlike F-test, determines whether a single variable is significant.
Size (X 6 ) does not significantly affect DER (Y).

The Effect of Domestic Ownership (X 2 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ) dan Management Expense (X 8 ) on DER (Y) a. Normality Test
A graphical display is used to summarize whether the data follow a normal distribution. Figure 4.6 P-P Plot of Normality Test The distribution of data points follows the normal reference line along the diagonal. This data distribution looks fairly normal, accordingly. b. Heteroscedasticity Test Figure 4.7 tests a regression model for heteroscedasticity by a graphical examination of the residuals. The residual scatterplot provides a visual examination of heteroscedasticity assumption and exhibits a random displacement of points with no clustering or systematic patterns. The points are also seen to be distributed above and below 0 (zero coordinate) on Y axis, indicating no signs of heteroscedasticity. This distribution satisfies the homoscedasticity assumption.

c. Multicollinearity Test
To indicate the extent to which multicollinearity is present, VIF is calculated for each predictor using SPSS with the following output: The VIF value for each predictor, as it appears in the table, is far below 10, i.e., X 2 = 1,172, X 5 = 1,081, X 6 = 1,174, X 7 = 1,059, and X 8 = 1,100. This suggests no multicollinearity is present as these predictors included in the model are not significantly correlated and, thus, are independent predictors. d. Autocorrelation Test Linear regression model is tested for autocorrelation that yields statistical value d = 0,784 in SPSS (14.0 for Windows).

e. Analysis of Multiple Linear Regression Equation
After all of the assumptions are checked, multiple linear regression analysis is run to examine the effect of the multiple X variables-Domestic Ownership (X 2 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) on Y variable-DER. This identifies a formula to make a prediction about the dependent variable based on the observed values of the independent variables in a causal relationship, i.e.: Y =  + b 2 X 2 + b 5 X 5 + + b 6 X 6 + b 7 X 7 + b 8 X 8 +  The resulting constant and regression coefficient can be used to formulate a linear regression equation: Y = 0,554 + 3,485 X 2 + 0,030 X 5 + 1,124 X 6 + 1,444 X 7 + 0,812 X 8 The equation is interpreted as follows: If Domestic Ownership (X 2 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) take on zero, DER (Y) will end up in 0,554 unit. b 2 = 3,485 If Domestic Ownership (X 2 ) increases by one unit and the others are held constant, DER (Y) will increase by 3,485 units. b 5 = 0,030 If Profitability (X 5 ) increases by one unit and the others are held constant, DER (Y) will increase by 0,030 unit. b 6 = 1,124 If Size (X 6 ) increases by one unit and the others are held constant, DER (Y) will increase by 1,124 units. b 7 = 1,444 If Credit Risk (NPL) (X 7 ) increases by one unit and the others are held constant, DER (Y) will increase by 1,444 units. b 8 = 0,812 If Management Expense (X 8 ) increases by one unit and the others are held constant, DER (Y) will increase by 0,812 unit. Table 4.14 presents the output of correlation coefficient estimation using SPSS statistics. Table 4.14 The Value of Product-Moment Correlation Coefficient a. Predictors: (Constant), Management Expense (X 8 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), Profitability (X 5 ), Domestic Ownership (X 2 ) b. Dependent Variable: DER (Y) The resulting value of correlation coefficient (r) is 0,410, which is interpreted based on the following objective criteria:

g. Simultaneous Hypothesis Test (F-Test)
Below is a set of simultaneously tested hypotheses.

h. Partial Hypothesis Test (T-Test)
T-test assesses a single regression coefficient at a time based on the hypotheses: 1) H 0 → b YX2 = 0 Domestic Ownership (X 2 ) does not significantly affect DER (Y).
The statistical test is T-test.
The T statistical value using SPSS is presented below: In the Anova output, t cal of X 2 = 3,372, X 5 = 0,431, X 6 = 1,140, X 7 = 1,043 and X 8 = 0,500. These values are compared to the probability distribution of the t value. For α = 5%, df (degree of freedom) = nk -1 = 140 -5 -1 = 134 in a two-tailed test, the resulting t Testing one variable at a time helps pinpoint which changes of Xs have an effect on Y based on those criteria with the following results; Domestic Ownership (X 2 ) significantly affects DER (Y) (3,372 > 11,978); Profitability (X 5 ) does not significantly affect DER (Y) (0,431 < 1,978); Size (X 6 ) does not significantly affect DER (Y) (1,140 < 1,978); Credit Risk (NPL) (X 7 ) does not significantly affect DER (Y) (1,043 < 1,978); and Management Expense (X 8 ) does not significantly affect DER (Y) (0,500 < 1,978). A plot of points that lie approximately on a straight line or scatter around the reference (regional) line indicates a normally-distributed set of data.

b. Heteroscedasticity Test
The nature of heteroscedasticity is examined using a graphical method below: Figure 4.13 Scatterplot of Heteroscedasticity The scatterplot exhibits no established patterns, and the data points lie above and below zero coordinate on Y axis. This indicates no heteroscedasticity of residuals, thus yielding homoscedastic data.

c. Multicollinearity Test
The following output indicates the VIF value for each free variable using SPSS statistics: The resulting VIF value for each free variable goes below 10, i.e., X 3 = 1,058, X 5 = 3,347, X 6 =1,377, X 7 =1,374, and X 8 =4,536. A VIF below 10 does not indicate high correlation among these free variables in the regression model, representing a linear combination of the independent variables.

d. Autocorrelation Test
The linear regression model is tested for autocorrelation that yields statistical value d = 0,999 in SPSS (14.0 for Windows).  ((1,29), the model is assumed to be positively autocorrelated. Figure 4.14 Zero-Order Autocorrelation Test When autocorrelation is problematic, the predictor variables are transformed (one time) using the estimate of ρ (rho) based on the d value in Durbin-Watson statistic (Gujarati, N. Damodar, Essentials of Econometrics, Second Edition, 1998: 394). Following the one-time variable transformation, autocorrelation test is rerun using SPSS (13.0 for Windows).

g. Simultaneous Hypothesis Test (F-Test)
Below is a set of simultaneously-tested hypotheses: There is no significant effect of Mixed Ownership (X 3 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ), and Management Expense (X 8 ) on DER (Y).
The statistical test is F-test. The F statistical value using SPSS is presented below: , Size (X 6 ), Profitability (X 5 ) b. Dependent Variable: DER (Y) In the Anova output, the resulting F cal is 4,361, which is compared to the probability distribution of Fvalue. At α=5%, db 1 (degree of freedom) = k = 5, and db 2 = nk -1 = 44 -5 -1 = 38, the resulting F  The F cal , as it appears in the curve, is greater than the F table (4,361 > 2,463). H 0 is therefore accepted, indicating that the group of Xs (Mixed Ownership, Profitability, Size, Credit Risk, and Management Expenses) is jointly significant in DER (Y).
The statistical test is T-test.
The t statistical value using SPSS is presented below: In the anova output, the resulting t cal of X 3 = 0,061, X 5 = 1,446, X 6 = -2,128, X 7 = 0,834 and X 8 = 3,207. These values are compared to the probability distribution of the T-value. At α = 5%, db (degree of freedom) = nk -1 = 44 -5 -1 = 38 in a two-tailed test, the resulting t tabel is 2,024 and -2,024. The partial test meets these underlying criteria: Reject H 0 in favor of H 1 if -t table ≥ t cal ≥ t table ; or Accept H 0 and, hence, reject H 1 if -t table < t cal < t table .
The partial t-test assesses, as Xs are not highly correlated, which X actually creates the effect on Y based on those criteria with the following results; Mixed Ownership (X 3 ) does not significantly affect DER (Y) (0,061 < 2,024); Profitability (X 5 ) does not significantly affect DER (Y) (1,446 < 2,024), Size (X 6 ) does

The Effect of Foreign Ownership (X 4 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ) and Management Expenses (X 8 ) on DER (Y) a. Normality Test
The graphical method below provides the examination of the data normality. Figure 4.17 P-P Plot of Normality Test The points on the plot align with the diagonal line, and, thus, the data set conforms to the normal distribution. b. Heteroscedasticity Test Figure 4.18 presents a graphical procedure to check for the potential heteroscedasticity in the application of regression analysis. Figure 4.18 Scatterplot of Heteroscedasticity The data points stray from the line in a non obvious fashion, with the distribution of points scattering randomly around zero on Y axis, thus no signs of heteroscedasticity. The homoscedasticity assumption of the regression model is therefore thoroughly verified for the predictive purposes. The resulting VIF for each free variable stays below 10, i.e., X 4 = 1,312, X 5 = 1,151, X 6 = 1,143, X 7 = 1,209, and X 8 = 1,106. A VIF below 10 indicates insignificant correlation among these variables, thus making them independent of each other.

d. Autocorrelation Test
The regression model is tested for autocorrelation in Durbin-Watson test. The resulting d statistic value is 1,615 in SPSS (14.0 for Windows).

g. Simultaneous Hypothesis Test (F-Test)
The simultaneously-tested hypotheses are as follows: There is no significant effect of Foreign Ownership (X 4 ), Profitability (X 5 ), Size (X 6 ), Credit Risk (NPL) (X 7 ) and Management Expense (X 8 ) on DER (Y).

h. Partial Hypothesis Test (T-Test)
T-test assesses a single regression coefficient at a time based on the hypotheses: The significance level is α = 5%.
The statistical test is T-test.
The T statistical value using SPSS is presented below: In the Anova output, t cal of X 4 = -0,097, X 5 = 3,533, X 6 = -0,359, X 7 = 0,727 and X 8 = -0,502. These values are compared to the probability distribution of the T-value. At α = 5%, db (degree of freedom) = n k -1 = 80 -5 -1 = 74 in a two-tailed test, the resulting t tabel is 1,993 and -1,993. The partial test meets these underlying criteria: Reject H 0 in favor of H 1 if -t table ≥ t cal ≥ t table ; or Accept H 0 and, hence, reject H 1 if -t table < t cal < t table .
The criteria predict Y on the basis of Xs with the following outcomes; Foreign Ownership (X 4 ) does not significantly affect DER (Y) (-0,097 > -1,993); Profitability (X 5 ) significantly affects DER (Y) (3,533 > 1,993); Size (X 6 ) does not significantly affect DER (Y) (-0,359 > -1,993); Credit Risk (NPL) (X 7 ) does not significantly affect DER (Y) (0,727 < 1,993); and Management Expense (X 8 ) significantly affects DER (Y) (-0,502 > -1,993). The result of hypothesis test confirms the insignificant partial effect of ownership structure on capital structure. The proportion of firm ownership does not measure the extent of debt instrument that allows financial latitude. Prior data reflect that firms take on debt financing more heavily over the years, and the ownership structure is bound to remain stable (Haruman, 2008) and (Imas, et al., 2015). Profitability has a weak effect on capital structure decision. Krishnan (1996), Badhuri (2002), Moh'd (1998), Majumdar (1999) (in Yuke and Hadri, 2005) and Imas, et al. (2015 point out that a firm which earns higher return on equity when its needs for external funding or debt decreases to fund new investment is able to earn at a higher rate than it pays for borrowed funds. A high-performance firm is expected to use its internal funds (retained earnings) and, thus, relies less on debt financing in its capital structure. The partial effect of firm size on capital structure also shows insignificant result. Rista and Bambang (2011), Heruman (2008) and Imas, et al. (2015 assert that a managerial decision that affects the financial condition of a firm is not greatly influenced by how much of total assets have been allocated among current and fixed assets. Consistent with Haruman (2008) and Imas, et al. (2015), the present study finds measuring and managing credit risk is of central importance for financial institutions and has no significant effect on the dynamic capital structure adjustment, notwithstanding. Exposure to credit risk across different firms varies widely. However, the tendency to take on a great deal of high-yield debt remains high. Management expense, as opposed to other previous variables, has major potential effect on the factors that influence the decisions concerning the capital structure. In accordance with Siringoringo (2012) and Imas, et al. (2015), the present study finds that relatively high management expenses commonly indicate an aggresive total cost associated with the increase in assets, thus exceeding the marginal costs of imposing a leverage ratio increase.

CONCLUSION
By considering the data of the entire conventional banks listed on IDX from 2013-2016, this present study empirically examines the effects of ownership structures and bank-specific characteristics on the capital structures. It has provided an in-depth understanding of firms' capital structure needs in a qualitative manner, highlighting the importance of evaluating how the capital structures help finance their assets, day-to-day operations and future growth. To this end, multiple linear regression is performed to gather and represent the predictive results concerning the correlation of capital structures and a number of variables. All hypotheses are confirmed insignificant, except one. The findings are statistically insignificant with respect to the relationship between ownership structure and capital structure; the relationship between profitability and capital structure; the relationship between firm size and capital structure; and the relationship between credit risk and capital structure. When it comes to management expenses, however, it can be ascertained that there is a significant relationship in the framework for evaluating the dynamic capital structure adjustment. These relationships can potentially affect a firm's financial decision and its adjustment and how firms are relying more heavily on the banking sector for their debt financing needs.

SUGGESTION
This study contributes to the extant literature on capital structures in banking institutions and fills the gap in the wide strand of literature by providing empirical evidence of the relationship between ownership structures and bank-specific characteristics in terms of how the observed firms manage their capital structures. There are different subjects of analysis in order to extrapolate key themes and results that help predict future trends, shed light on previously hidden disciplinary pathways that can be applied to practice and provide means for understanding relevant pivotal research issues based on research approaches appropriate for the development of knowledge in a given study. In addition, the author suggests these specific aspects be observed in more depth: