“The effect of the government bond value on the intermediary function of banks in the capital market of Indonesia”

ARTICLE INFO Rosemarie Sutjiati Njotoprajitno, Bram Hadianto and Melvin (2020). The effect of the government bond value on the intermediary function of banks in the capital market of Indonesia. Banks and Bank Systems, 15(3), 199-206. doi:10.21511/bbs.15(3).2020.17 DOI http://dx.doi.org/10.21511/bbs.15(3).2020.17 RELEASED ON Wednesday, 07 October 2020 RECEIVED ON Thursday, 14 May 2020 ACCEPTED ON Tuesday, 01 September 2020


INTRODUCTION
The government of Indonesia has already issued and managed bonds to finance its budget following the Minister of Finance Decree No. 101/ KMK.017/2000. This circumstance has opened the opportunity for its citizens to join the national development and get the coupon regularly paid for the issued bonds that they have bought (Law No. 24 of 2002). Two arguments exist regarding the issuance of government bonds and the effect. First, the supporting one, as Abbas and Christensen (2007) demonstrate. They argue that by selling the bonds, the government can create macroeconomic stability, for example, reducing inflation and protecting the state from monetary crisis and other external crises. Second, the contra one, as shown by Hanson (2007). He declares that although domestic debt can help the capital market, the large amount of this debt has a similar risk to foreign debt.
Additionally, the issuance of government bonds can be a competition for banks to search for funds to redistribute them back to society in a loan. Consequently, what the government executes disturbs the intermediary function of commercial banks. This condition was supported by Christensen (2005) declaring the government debt goes down the bank lending to private sectors; DeBonis and Stacchini (2013) pointing out that the debt issued by the government drops the bank credit growth. Based on their study, Altaylıgil and Akkay (2013) suggest that the government should reduce its debt to facilitate the banks to distribute credits to the private sectors to enhance economic growth. Correspondingly, Anyanwu, Gan, and Hu (2017) confirm that government debt diminishes bank credit to private sectors. Similarly, Mwakalila (2020) shows that domestic government liabilities tend to cut bank credit.
Unfortunately, this crowding-out hypothesis has been broken by Utari, Kurniati, and Surjaningsih (2011), stating that government debt has a positive impact on a bank's ability to distribute funds; Akpansung (2018), announcing no effect; and Benayed and Gapsi (2020), finding that the inverted-U shape curve exists. Furthermore, they explain that under the starting point of 52% of GDP, domestic public debt supports private banks for lending money. Upper this point, bank credits to the private sector fall.
This study also utilizes bank size and bad loans as the control variables. These variables are used because they become the determinants of the bank in channeling funds. This research intends to ensure the crowding-out phenomenon by investigating and analyzing the government bond effect on the bank intermediary function in the Indonesia capital market by employing its size and loans as a control variable.

THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT
By issuing bonds to finance the state budget deficit, the government attracts the attention of the public to allocating money in this instrument. This condition makes money flow from the corporate bond market (Wibowo, Passagi, and Prasetyo, 2018) and the banking industry (Wibowo, 2018) to the government bond market (

Research variables and their measurement
The variables in this study have two positions: outcome and explanatory. A bank intermediary function acts as an outcome. On the other hand, government bond value, bank size, and bad loans are explanatory variables. Table 1 shows the measurement of these variables.

Population and samples
The   Likewise, we employ the simple random sampling method to take 23 banks as the samples. Assuring the randomness, we also utilize the random number generated by Microsoft Excel, as Hartono (2012) explains. After that, the names of the banks obtained are available in Table 2.

Data analysis method
Referring to the variable scale in Table 1, this study uses the regression with pooled data adopting the ordinary least square method as the parameter estimation method (Nachrowi & Usman, 2006). (Nachrowi and Usman, 2006). Additionally, this model can be seen in equation (2).
The regression model must reach the test of some classical assumptions to yield the best, linear and unbiased estimators (BLUE). In other words, this model's errors have to follow the normal distribution and be free from the impact of explanatory variables (the absence of heteroscedasticity). Additionally, there is no significant correlation between the independent variables (the nonappearance of multicollinearity).
The residuals have to be random (the absence of autocorrelation) (Ghozali, 2016).
• To attest to the normality of residuals, the Kolmogorov-Smirnov test was used. The normality happens when the asymptotic significance of the Z-statistic of K-S is higher than a significance level (α) of 5% or a restricted one of 1%.
• To prove the absence of heteroscedasticity, the Glesjer test was used. Heteroscedasticity does not occur when the probability of the t-statistic for all independent variables is higher than a significance level (α) of 5%.
• To ascertain the absence of autocorrelation, the runs test based on the mode was used. Autocorrelation does not exist when the asymptotic significance of Z-statistic is higher than α of 5%.
• To detect the nonexistence of multicollinearity, the variance inflation factor (VIF) was compared with the cut-off point of 10. Multicollinearity does not exist when the VIF of each independent variable is lower than 10.
Furthermore, to examine the regression coefficients, β 1 , β 2 , and β 3 , t-statistic is used by comparing its probability with the significance level by indicating this following hint: • If the probability of t-statistic is less than α of 5%, the null hypothesis is declined.
• If the probability of t-statistic is above or the same as α of %5, the null hypothesis is acknowledged.

RESULTS AND DISCUSSION
This research employs 23 banks with a 9-year lifespan, bringing the total number of observations to 207. Moreover, 207 related to four variables was statistically described in Table 3.
•  Table 4 shows the examining results of classical assumptions with the explanation as follows: • For the normality test outcome, the asymptotic significance (2-tailed) of the Z-statistic of KS is 0.031. Since this value exceeds the delimited significance level of 1%, errors trace the normal distribution.
• For the heteroscedasticity test outcome, the probability of the t-statistic of LNGBV, LNTA, and G_NPL is 0.092, 0.174, and 0.116, respectively. Since each value exceeds the significance level (α) of 5%, the absolute error is not affected by LNGBV, LNTA, and GN_PL. In other words, there is no heteroscedasticity.
• For the multicollinearity detection outcome, the values of VIF for LNGBV, LNTA, and G_ NPL are 1.068, 1.172, and 1.116, respectively. Since each value exceeds 10, multicollinearity does not exist in the regression model.
• For the autocorrelation test outcome, the asymptotic significance (2-tailed) of the Z-statistic of KS shows 0.921. Since this value exceeds α of 5%, errors are random. As a consequence, there is no autocorrelation problem in the regression model. Table 5 presents the regression model's estimation outcome with pooling data and shows the probability of the t-statistic of 0.0034, 0.0001, and 0.0092 for each regression coefficient, LNGBV, LNTA, and NPL, to examine the null hypothesis. Since the probability of the t-statistic is lower than α of 5%, this study discards all the null hypotheses. This circumstance means that research hypotheses 1b, 2, and 3 are recognized due to a positive regression coefficient.  By showing a positive effect of government bonds on bank intermediary function, the crowding-out does not exist. The government does not need to worry about that because the commercial banks can use this opportunity by buying bonds as compensation to cover credit risk. Additionally, large banks tend to distribute more funds than small banks because they can manage the risk by forming their asset portfolios. This implies the certification of the banking officers in large banks is essential to guarantee the quality of risk management and governance. Moreover, the liquidity problem tends to belong to banks with high bad loans. To fix this problem, restructuring loans for their borrowers can become an alternative.

CONCLUSION
This paper explores the impact of the government bond value on the bank intermediary function with the samples and the data related to the research variables from the banks listed on the capital market of Indonesia between 2010 and 2018. This study infers that the government's bonds can increase the banks' channeling funds because the crowding out does not exist. As a control variable, bank size and bad loans possess a positive influence on this function.
Although three explanatory variables significantly affect bank intermediary function, this study still has some limitations, i.e., the number of explanatory variables used and the population's scope.
• Concerning the first limitation, the next scholars can add internal and external bank factors as other explanatory variables in their research model to overcome it. Examples of internal factors are the interest rate of bank deposits and loans, profitability, bank capital, capital adequacy ratio, operating expense to revenue ratio, and growth of total deposits. Meanwhile, examples of external factors are gross domestic product, inflation, unemployment, and economic development.