FACTORS AFFECTING PERFORMANCE AND THE IMPACT ON CUSTOMER TRUST OF SYSTEMIC BANK IN INDONESIA PERIOD 2012-2017

The aim of this study was to analyze the factors that influence Performance and Its Impact on Customer Trust of Systemic Bank period 2012-2017. The independent variable of this study uses the financial ratio of Non-Performing Loans (NPL), Cost Income Ratio (CIR), Net Interest Margin (NIM), and Capital Adequacy Ratio (CAR). Performance measured by Return on Assets (ROA), and for the Customer Trust variable measured by Bank Rating. The study used panel data analysis and the process was using Eviews version 9.0. The results of the analysis in this study indicate that (1) NPL has a negative and not significant effect on ROA, (2) CIR has a negative and significant effect on ROA, (3) NIM has a positive and significant effect on ROA, (4). CAR has a negative and not significant effect on ROA, (5) NPL, CIR, NIM and CAR simultaneously have a positive and significant effect on ROA, (6) NPL has a positive and not significant effect on Rating. (7) CIR has a negative and not significant effect on Rating, (8) NIM has a negative and significant effect on Rating. (9) CAR has a positive and not significant effect on Rating, (10) NPL, CIR, NIM and CAR simultaneously have a positive and significant effect on Rating, (11) ROA has a negative and not significant effect on Rating.


Introduction
The economic crisis in 1998 and the global crisis in 2008 affected banks experiencing a decrease operating profits, so those banks have to maintain their existence in national banking. This profit decrease caused by the higher of costs of funds, the other of source bank losses is foreign exchange transactions, especially impairment exchange rate of rupiah to US dollars, increasing of bad debt, the difficulties of liquidity, etc. It can affect potential of financial losses and it possible to cause the bank bankruptcy so the bank is liquidated. According to Bank Indonesia data in 1998, the depreciation of the rupiah reached 600% in a period of less than 1 year, which is from Rp2.350 to Rp16.000 each USD. Meanwhile, interest rates of inter-bank had reached 60% per year. The health rating system using the Risk-Based Bank Rating (RBBR) method is often referred to as a risk-based system. In this RBBR concept, banks are required to maintain and improve bank health by applying prudential principles and implementing risk management in carrying out business activities. Bank Indonesia states that the implementation of RBBR is more of an analysis and judgment aspect where there is an effort to develop a more flexible level rating system. After obtaining ratings from risk profiles, GCG ratings, earnings quality, and capital quality, a rating can be obtained from the bank's soundness level, namely by giving weight to each of these ratings and calculating bank ratings. In 2008, a decrease in the prices level for agricultural, energy and housing commodities, accompanied by an increase in oil prices to more than $ 100 per barrel, caused many banks to experience a decline in profits. In contrast, in 2009 when interest rates continued to decline, bank profits continued to skyrocket to date (IBI, 2016: 145) Based on Indonesian Banking Statistics data, the performance of conventional commercial banks (consisting of systemic banks and non-systemic banks) that were measured using ROA from 2005 to 2014 increased and tended to be stable. Only in 2008 did it decline due to the global economic crisis which affected the Indonesian economy. The performance of conventional banks in the last 3 years has decreased significantly even ROA in 2015 and 2016 was below the 2008 ROA during the global economic crisis.
The phenomenon that occurs in systemic banks is even though the performance has decreased but the level of customer confidence measured by the rating according to Pefindo tends to be stable and even increase. This is because the systemic bank conditions are resilient despite an economic crisis which if a systemic bank experiences failure will have an impact on other banks.

Return on Assets
According to Kasmir (2009: 197), several ways to measure the performance of a company, namely: 1. Gross profit margin (GPM). This measurement is a measure of the percentage of each sale after the company pays the cost of goods sold. The higher the gross profit margin the better. 2. Operating profit margin (OPM). This measurement is a measure of the percentage of each remaining sale after all other costs and expenses are deducted except interest and taxes. 3. Net profit margin (NPM). This measurement is a measure to measure the percentage of company profits after deducting all costs from expenses including interest and taxes. 4. Return on assets (ROA). This measurement is a measure of the effectiveness of management in generating profits with available assets. 5. Return on equity (ROE). This measurement is a measure of the return the owner has obtained for investment in the company.
To evaluate the performance of the bank in obtaining profits, the bank's financial ratios can be calculated from financial statements compared to the performance in the past period, and the current performance of competitors. Return on Assets (ROA), net income divided by total assets, [Salim et. al [29] is a ratio that is often used to measure bank performance. ROA depends on the ability of the bank to obtain interest income, control costs, interest and operational efficiency, and others. By looking at the ROA component, bank management can concentrate on problems that make banks unable to produce a good performance. From ROA, banks can continue research to find the root causes of declining performance of banks in generating profits (Indonesian Bankers Association, 2016: 145)

Rating
According to Manurung (2008), a rating is one of the variables considered by investors when deciding to invest in a company. The information contained in the rating will show the extent to which the ability of a company to pay its obligations on funds invested by investors. Companies that have high ratings are usually preferred by investors compared to companies that have very low ratings. Therefore, so that the bonds of a company that has a fairly low rating, can be sold on the market, then usually investors will determine a higher premium, as a compensation for the risks borne by investors.

Systemic Bank
A Systemic Bank is a Bank that is due to the size of assets, capital and liabilities; network size or complexity of transactions for banking services; and the relationship with other financial sectors may result in the failure of part or all of the other Bank or financial services sector, both operationally and financially, if the Bank is experiencing a failure or failure. Capital Surcharge is additional capital that serves to reduce the negative impact on financial system and economic stability in the event of a Systemic Bank failure through increasing the Bank's ability to absorb losses. Banks designated as Systemic Banks must establish a Capital Surcharge

Research Design
The design of this study uses quantitative research. According to Sugiyono (2010: 11), quantitative research sees the relationship of variables to the object under study more causal, so that in the study there are independent and dependent variables. Thus, this study can be categorized into types or quantitative causality research designs.

Population and Samples
The research population is a commercial bank registered as a systemic bank in Indonesia. The sample selection uses a purposive sampling technique, namely the technique of determining samples with certain considerations. This is so that the sample used has predetermined criteria. Banks that are consistently included in systemic banks in the period 2012-2017 will be the sample of this study.

Descriptive Statistics Analysis
Descriptive Statistics is a statistic that is used only to describe the patterns or conditions of a sample of a population. The analytical tool used to describe these patterns is usually illustrated through central tendency estimates, such as mean values, median values, and frequently occurring values

Data Analysis Method
The analytical method used in this study is panel data analysis. Panel data analysis is a combination of both types of time series data and cross-section, so that the data held is not only based on one period but the data covers several periods. According to Salim et al (2018: 3) can use inferential statistics to test hypotheses about a population based on the sample as the subject of research. The testing in this study uses the program eviews 9.0.
The estimation method of panel data regression models are three commonly used approaches, namely:

Common Effect Model
The Common Effect Model is the simplest panel data approach. In this approach, it does not pay attention to time and individual dimensions, so that the treatment of data between companies is assumed to be the same in various time periods. This model only combines time series data and cross section in the form of a pool, estimating it using a small square / pooled least square approach.

Fixed Effect Model
Fixed Effect Model assumes that there are different effects between individuals. This difference can be accommodated through differences in the intercepts. The Fixed Effect Model technique is a technique of estimating panel data by using a dummy variable to capture intercept differences. The definition of Fixed Effect is based on the difference in interception between companies but the intercept is the same between times. In addition, this model also assumes that the regression coefficient (slope) remains between companies and between times.

Random Effect Model
Random Effect Model is a panel data regression estimation model with the assumption that the constant slope coefficient and intercept are different between individuals and between times. Random Effect Model estimates panel data where interference variables may be interconnected between time and between individuals.
Selection of the panel data regression model can be done using the Chow test, Hausman test and Lagrange Multiplier Test. The chow test is a test to determine a fixed effect model or pool that is most appropriate to use in estimating panel data. The Chow test in this study uses program eviews.
Hausman test aims to determine whether the model is fixed effect or random effect. The fixed effect model assumes that the independent variable correlates with the error, while the random effect is the opposite. The panel data model with fixed effect is estimated by OLS (Ordinary Least Square), while random effects are estimated by GLS (Generalized Least Square). Lagrange Multiplier, to find out whether the Random Effect model is better than the Common Effect model. The following is a research framework:

Results and Discussions
This study is to examine the effect of Non Performing Loans, Cost Income Ratio, Net Interest Margin and Capital Adequacy Ratio by statistical testing in order to find out the four variables on performance and their impact on systemic bank customer confidence levels for the period 2012-2017. During this period, there were 11 banks that were consistently included in the systemic bank category: BRI, Mandiri, BCA, BNI, CIMB Niaga, BTN, Panin, Maybank/BII, OCBC NISP, Danamon and Permata  [32] 2) The Hausman test produces p-value (Prob) of 0.8914 (more than 0.05), so we can reject H0 and accept H1. This shows that the random effects model method is more appropriate to use than the fix effects model method 3) The Lagrange Multiplier test produces p-value (Prob) of 0.0293 (less than 0.05), so we can reject H0 and accept H1. This shows that the random effects model is the best model 4) The Constant value (a) of 7.295867 means that if the variables in this study are NPL, CIR, NIM and CAR worth 0, then ROA (Y) is obtained at 7.295867. 5) The variable NPL regression coefficient value obtained at -0.050574 negative value indicates the existence of an opposite direction relationship between the ROA variable and NPL, which means that if the NPL increases by 1 unit then ROA will decrease by 0.050574 assuming that other independent variables remain. 6) CIR variable regression coefficient of -0.083602 negative value indicates the existence of an opposite direction relationship between the ROA variable and CIR, which means if CIR increases by 1 unit then ROA will decrease by 0.083602 assuming that the other independent variables remain. 7) NIM variable regression coefficient of 0.319315 positive value indicates a unidirectional relationship between the ROA variable and NIM, which means that if the NIM increases by 1 unit then ROA will increase by 0.319315 assuming that the other independent variables remain. 8) The variable regression coefficient value of -0.006923 negative value indicates the existence of an opposite direction relationship between the ROA variable and CAR, which means if the CAR has increased by 1 unit then the ROA will decrease by 0.006923 assuming that the other independent variables remain. 9) Based on the F test, it is known that the F-Statistic value is 298.96 with a significant level of 0,000. This shows that NPL, CIR, NIM, and CAR simultaneously influence ROA 10) The determination coefficient is 0.9514. This shows the NPL variable, CIR, NIM, and CAR simultaneously able to provide an explanation on the ROA variable of 95.1467% while the remaining 4.8533% is explained by other factors not examined in this study.  [33] Based on the table above, the following is an explanation of the research equation: 1) Based on the Chow Test it is known that p-value (Prob) is equal to 0,000 (less than 0.05), so we can reject H0 and accept H1. This shows that the fixed effects model is more appropriate to use than the common effects model 2) The Haussman test produces p-value (Prob) of 0.8914 (more than 0.05), so we can accept H0 and reject H1. This shows that the fixed effects model is the best model 3) Constant value (a) of 107.3155 means that if the variables in this study are NPL, CIR, NIM and CAR is 0, then a rating of 107.3155 is obtained. 4) NPL variable regression coefficient obtained at 0.328536 positive value indicates a unidirectional relationship between the Rating variable and NPL, which means that if the NPL increases by 1 unit, the Rating will increase by 0.328536 assuming that the other independent variables remain. 5) CIR variable regression coefficient of -0.007145 negative value indicates the existence of an opposite direction relationship between Rating variable with CIR, which means if CIR increases by 1 unit then Rating will decrease by 0.007145 assuming that other independent variables remain. 6) The NIM variable regression coefficient of -2.386714 negative value indicates that there is an opposite direction between the Rating variable and NIM, which means that if the NIM increases by 1 unit, the Rating will decrease by 2.386714 assuming that the other independent variables remain. 7) CAR variable regression coefficient of 0.202912 positive value indicates the existence of a unidirectional relationship between the Rating variable and CAR, which means if the CAR has increased by 1 unit then Rating will increase by 0.202912 assuming that the other independent variables remain. 8) Based on the F test, the F-Statistics value is known to be 15,993 with a significant level of 0,000. This shows that NPL, CIR, NIM, and CAR simultaneously influence Rating 9) The determination coefficient is 0.814. This shows the NPL variable, CIR, NIM, and CAR simultaneously able to provide an explanation for the Rating variable of 81.4% while the remaining 18.6% is explained by other factors not examined in this study.  [34] Based on the table above, the following is an explanation of the research equation: 1) Based on the Chow Test it is known that p-value (Prob) is 0.0000 (less than 0.05), so we can reject H0 and accept H1. This shows that the fixed effects model is more appropriate to use than the common effects model 2) The Haussman test produces p-value (Prob) of 0.0710 (more than 0.05), so we can reject H0 and accept H1. This shows that the random effects model method is more appropriate to use than the fix effects model method 3) The Lagrange Multiplier test produces p-value (Prob) of 0.0000 (less than 0.05), so we can reject H0 and accept H1. This shows that the random effects model is the best model 4) Constant value (a) of 98.85600 means that if the variable in this study is a value of ROA 0, then a rating of 98.85600 is obtained. 5) ROA variable regression coefficient obtained at -0.530646 negative value indicates the existence of an opposite direction relationship between the variable Rating with ROA, which means that if ROA increases by 1 unit, the Rating will decrease by 0.530646 assuming that the other independent variables remain. 6) The determination coefficient is 0.0456. This shows that the ROA variable is able to provide an explanation for the Rating variable of 4.56% while the remaining 95.44% is explained by other factors not examined in this study. Effect of NPL on ROA Ismail (2013: 127) explains that the impact of problem loans is a decrease in profits. Decreasing profits will have an impact on decreasing return on assets (ROA). The results of the above research can be concluded that the higher the non performing loan will reduce the systemic bank performance but not significant because the level is significantly greater than 5%. Systemic banks that are generally able to maintain credit quality in the healthy category. This is evidenced by the mean NPL ratio at 2.45%. This does not support the study of Petria Dendawijaya (2009: 120) explains that CIR is the ratio of operational costs used to measure the level of efficiency and ability of banks to carry out their operations. Big CIR lack of efficiency will result in lower profits. The results of this study support the theory and also previous studies conducted by Mathuva (2009) Riyadi (2008: 135) explains that high net interest income will result in increased profit before tax so that ROA also increases ". Based on the results of this study, it is known that NIM has a significant positive effect. This supports previous studies conducted by Khan

Effect of CAR on ROA
The Bank's capital is primarily intended to cover potential unexpected losses and as a reserve in the event of a banking crisis. Risk Exposures that accumulate due to banks carrying out activities need to be supported by high-quality capital. In the event of a global crisis, banks generally suffer losses from the credit business and require actions to eliminate loans that have an impact on capital (IBI, 2016) This study does not support previous research conducted by Boadi  and Alper and Anbar (2011). The regulator sets a minimum ratio of 8% CAR. Banks in general will always maintain and even increase their CAR ratios each year to be able to expand their business freely. The average Systemic bank CAR ratio is 17.6% far above the 8% stipulation.
Effect of NPL, CIR, NIM, and CAR on ROA NPL variables, CIR, NIM, and CAR are simultaneously able to provide an explanation for the variable ROA of 95.1467% while the remaining 4.8533% is explained by other factors not examined in this study. The magnitude of the coefficient of determination means that the level of the relationship between NPL, CIR, NIM and CAR to ROA is very strong. This study supports previous research conducted by Ariyanti et al (2017) and Maudhita and Thamrin (2018). This is also in accordance with the theory that risk-based bank health management can improve bank performance.

Effect of NPL on Rating
Based on the calculation of panel data analysis for NPL variables does not significantly influence rating with a positive relationship direction at a significant level of 0.05 (5%) or in other words, the NPL variable has a positive and not significant effect on the rating at the 95% confidence level. This study does not support previous research conducted by Lestari and Indriyani (2016). This is because the research sample is a bank with the largest asset size in Indonesia, which even though NPL increases it tends to be unhealthy but still has a good rating and is worthy of investment. This is evidenced by the gem bank NPL of 8.83% in 2016 but has an AAA or 100 rating on the ratio scale (Prime).

Effect of CIR on Rating
Based on the calculation of panel data analysis for the CIR variable it has a negative and not significant effect on the rating at the 95% confidence level. This study supports previous research conducted by Lestari and Indriyani (2016) and Pramana and Yunita (2015). This is because the research sample is a bank with the largest asset size in Indonesia, which although CIR is inefficient but still has a high level of customer confidence. This is evidenced by the gem bank CIR of 150.77% in 2016 but has an AAA or 100 rating on the ratio scale (Prime).

Effect of NIM on Rating
Based on the calculation of panel data analysis for the NIM variable it has a negative and significant effect on the rating at the 95% confidence level. This study does not support the research conducted by Pramana and Yunita (2015). Based on descriptive statistics, the NIM ratio at Maybank banks in 2012 declined in 2013 but the rating of Maybank's bank actually increased. Thus, although the ability of earning assets to generate net interest income declined significantly, the systemic bank customer confidence level continued to increase.

Influence of CAR on Rating
From the calculation of panel data analysis for the CAR variable has a positive and not significant effect on the rating at the 95% confidence level. This study supports previous research conducted by Pramana and Yunita (2015). This is because the research sample is a bank with the largest asset size in Indonesia, which in the CAR ratio and rating is in a stable position.

Effect of NPL, CIR, NIM, and CAR on Rating
Based on the results of the simultaneous test it can be concluded that NPL, CIR, NIM and CAR simultaneously have a significant effect on rating. NPL variables, CIR, NIM, and CAR simultaneously were able to explain the rating variable at 81.4480% while the remaining 18.552% was explained by other factors not examined in this study. With the magnitude of the determination coefficient of 81.4480%, it means that the level of the relationship between NPL, CIR, NIM and CAR to the rating is very strong. This is consistent with the research conducted by Pramana and Yunita (2015).

Effect of ROA on Rating
Based on the results of the partial influence test it can be concluded that ROA does not significantly influence rating with a negative relationship direction at a significant level of 0.05 (5%) or in other words, the ROA variable has a negative and not significant effect on the rating at the 95% confidence level. The ROA variable is able to provide an explanation on the rating variable of 4.5616% while the remaining 95.4384% is explained by other factors not examined in this study. With the magnitude of the determination coefficient of 4.5616%, it means that the level of relationship between ROA and rating is very weak. But what's interesting is that the relationship between ROA and Rating is negative, this shows that even though bank performance is declining, the level of customer confidence is increasing. This is different from the research conducted by Pramana and Yunita (2015) which explains that ROA has a not significant positive relationship and also Lestari and Indriyani (2016) research which states that ROA has a significant positive relationship to rating.

Conclusions and Recommendations
Based on the result of the study, the researcher conclude the factors that Affect Performance and the Impact to Systemic Bank Customer Trust as follows: 1) Non Performing Loans (NPL) have no effect on Return on Assets (ROA) The average systemic bank NPL of 2.45% is considered fine, so it does not affect the performance of systemic banks, 2) Cost Income Ratio (CIR) has a negative and significant effect on Return on Assets (ROA). If the ratio is getting smaller means the more efficient operational costs incurred by the bank so that it can improve systemic bank performance, 3) Net Interest Margin (NIM) has a significant positive effect on Return on Assets (ROA). The ability of productive assets to generate net interest income has been very good. Increasing productive assets increases the interest income and systemic bank performance, 4) Capital Adequacy Ratio (CAR) does not affect Return on Assets (ROA). The average systemic bank CAR of 17.6% is classified as very well so it does not affect the systemic bank performance. 5) NPL, CIR, NIM and CAR simultaneously have a significant positive effect on ROA. This proves that the Risk Based Bank Rating financial ratios can affect the performance of systemic banks, 6) Non Performing Loans (NPL) has no effect on Rating. The average systemic bank NPL of 2.45% is classified as well so it does not affect the level of customer trust, 7) Cost Income Ratio (CIR) does not affect Rating. CIR on systemic banks affects performance but does not affect the level of customer trust, 8) Net Interest Margin (NIM) has a significant negative effect on Rating. So even though the ability of earning assets to generate net interest income decreases, the systemic bank customer trust level keep rising, 9) Capital Adequacy Ratio (CAR) does not affect Rating. The average systemic bank CAR of 17.6% is classified as very well so it does not affect the level of customer trust, 10) NPL, CIR, NIM and CAR simultaneously have a positive effect on Rating. This proves that the Risk Based Bank Rating financial ratios can affect the systemic bank customer trust level, 11) Return on Assets (ROA) has no significant negative effect on Rating. So that even though performance is decreasing, systemic bank customer trust keep rising.