Predicting Volatility of Non-Performing Financing: Lessons from Indonesian Islamic Banking Industry

Islamic banking financing is carried out into three categories, namely financing based on economic sectors, financing based on contracts, and financing based on types of use. The funding is faced with risks where the customer is unable to pay the loan. This study discusses the forecasting of Islamic banking problem financing in Indonesia using monthly data from 2003 to 2019. Prediction is made using the Autoregressive Conditional HeteroscedasticityGeneralized Autoregressive Conditional Heteroscedasticity (ARCH-GARCH) method. Forecasting from the results of this study shows that non-performing financing tends to decrease in Indonesia. Islamic banking has been able to manage problem financing well. Islamic banking financing should be focused on sectors that have low risk. The results of this study are undoubtedly useful for stakeholders to make policies to improve the quality of the funding.


INTRODUCTION
Banking as an intermediary institution that connects investors with those who need capital has a strategic role in the economy of a country (Levine, 1997). The development of the banking world, especially sharia banking will surely face challenges that must be responded appropriately (Budiono, 2017). One of the challenges is that there are changes in world economic conditions that can have an impact on bank financing (Supartoyo, Juanda, Firdaus, & Effendi, 2018).
Islamic banking has gained traction in Indonesia, although the market share of Islamic banking is still small but the potential for growth continues to exist considering that Indonesia is the largest Muslim population in the world with various challenges (Rizvi, Narayan, Sakti, & Syarifuddin, 2019). One of the challenges is that financing carried out by Islamic banking will be faced with conditions where the capital borrower is no longer able to pay bills. Nonperforming financing is defined as loans that are past due and cannot be repaid (Bolognesi, Stucchi, & Miani, 2019). In Islamic banking, the problematic financing indicator is reflected in the problem financing with non-performing financing (NPF) indicator.
Research related to NPL/NPF has been conducted by Li and Zhang (2019) analyzing related to credit options and risks in Hong Kong. Zhang, Cai, Dickinson, and Kutan (2016) who examined the impact of NPLs on bank behavior in China. Tsai, Lu, and Hung (2016) investigated information in newspapers influencing credit risk in the United States. Gatfaoui (2017) investigates the relationship between credit and equity markets in the United States. Chang, Fuh, and Kao (2017) ;Cucinelli, Battista, Marchese, and Nieri (2018); Binici and Hutchison (2018); Ashour and Hao (2019) have improved the credit risk of companies in China and Western Europe.
Bian, Lin, and Liu (2018) study credit risk with property prices in China. Dendramis, Tzavalis, and Adraktas (2018) investigated economic recession, pressures on the financial sector and political instability are factors that influence problematic financing. Zamore, Beisland, and Mersland (2019) analyze the relationship between geographical diversification and credit risk in microfinance. Niu and Hua (2019) studied the optimal contract between managers and shareholders in the credit risk model and the impact of moral hazard on corporate credit risk. Aysan and Disli (2019) examine whether there is a causal relationship between loans and problem loans in Turkey.
From the existing research, most of them discuss the effect of problem financing and none have discussed the forecasting of problem financing. Therefore, this research was conducted to fill the gap of existing research. Thus, the purpose of this study is focused on discussing forecasting problematic financing. This research has important implications for risk 3 managers and investors for optimal investment strategies (Troster, Tiwari, Shahbaz, & Macedo, 2019). This research is also deemed necessary because it can contribute to stakeholders to determine policy alternatives.

METHOD
The data used in this study are secondary data from non-performing loans (NPF) of 14 sharia commercial banks and 20 sharia business units monthly from January 2003 to November 2019. Sharia banking sector data is random (random) and has high volatility, causing variable that is not constant. Volatility modeling is very important for risk management (Caporale & Zekokh, 2019). Therefore, we need a model that can be used to predict problem financing with the condition that there is heteroscedasticity (McCauley, 2009).
The first step taken in this study is the identification of a model. The identification of this model is intended to find out what model will be used in research. So that data can be modeled with Autoregressive Moving Average (ARMA) or Autoregressive Integrated Moving Average (ARIMA), the thing that needs to be considered is that data must be stationary. If the stationary data is at a level, it can be modeled with ARMA. However, if the stationary data is at a difference, it can be modeled with ARIMA.
After getting data that has been stationary, then plots are made in the form of Autocorrelation (AC) and Partial Correlation (PAC) that are used to determine the order of the alleged model. Furthermore, after obtaining the alleged model, significance testing, residual testing and normality testing are carried out. Whereas in selecting the best model, it is chosen based on the value of the smallest Akaike Info Criterion (AIC) and the largest log likelihood. The next step is to test heteroscedasticity in the residual model. If there is an element of heteroscedasticity, modeling is done using Autoregressive Conditional Heteroscedasticity Based on Figure 1, the average NPF value throughout the observation was 3.9 percent, with the highest value of 6.6 percent and the lowest value of 2.2 percent. Data distribution in general is normally distributed. This condition can be seen in the Jarque-Bera probability value which is above 5 percent while the symmetrical distribution indicated by the variable generally has a long tail on the right side (positive skewness). The kurtosis that appears, it turns out that the height measure of the data distribution is generally (leptokurtic). Then the results from standard deviations indicate the distribution of data in the sample by how close the individual data points are to an average of 0.96 percent. The downward trend in non-performing financing in Islamic banking shows an improvement in the performance of Islamic banking in managing to finance disbursed to debtors. The decline in this trend is inseparable from the role of regulators in issuing policies that support economic growth so that financing disbursed always increases and nonperforming financing decreases. Islamic banking also avoids channeling financing to sectors that have high risk industries with high risks and are not in accordance with existing regulations in Islamic banking. This was done to create a sustainable financing climate.
Based on the results of the stationary data testing using the Augmented Dickey-Fuller (ADF) test, the NPF shows that the data is not stationary at the level (prob * 0.136> 0.05), but the data has been stationary at the difference (prob * 0,000 <0.05). As for using images there is a plot between observations with the time when the data level is not stationary and when the difference data is stationary by having a constant average and variance ( Figure 2).  (Table 1).   (24) The selection of the best ARIMA models includes the ARCH model, the GARCH model, the ARCH in Mean (ARCH-M) model, the ARCH Threshold model (TARCH), and the Expansion GARCH (EGARCH) model. Based on the results of the processing that has been done, it can be concluded that the best model is the TARCH model based on the largest log likelihood value and the smallest AIC (Table 2). Furthermore, after getting the best model, ARCH-LM will be tested for the best model. Based on the results of the processing that has been done, it can be seen that the value of Prob.
Chi-Square (1) is 0.5074, where the value is greater than the value of 0.05 so it can be concluded that the estimated TARCH model is free from the ARCH effect (Table 3).   This study predicts the movement of NPF of Islamic banking over the next 22 months.
From the results of forecasting that have been done, it can be seen that the tendency of NPF has decreased every month with an average decrease of 6.92 percent. Forecasting is done in early 2020, which has increased from the actual data that has been obtained. November 2019 initially NPF value of 3.18 percent to 4.09 percent in January 2020. Then in the following month NPF of Islamic banking decreased with until the end of forecasting, namely December 2021.
Forecasting the movement of the NPF which tends to fall will have an impact on the increase of financing carried out by Islamic banking, especially to sectors that have low risk (Chen & Lin, 2016). The characteristics of companies which include size, sector and legal status are common obstacles faced by banks in providing financing (Mertzanis, 2016).

Significance Parameters (α = 5%) Log
Likelihood AIC Figure 3. Dynamic Forecast NPF forecasting of sharia banking, which tends to decline is inseparable from the efforts of sharia banks to implement rescheduling, restructuring, and reconditioning programs (Ashraf, Zheng, Jiang, & Qian, 2020). On the other hand, Islamic banking has also made intensive underwriting and monitoring processes for financing customers and has been more careful in expanding financing (Hassan, Khan, & Paltrinieri, 2019). One of the efforts is by entering into financing with segments, sectors, and customers with a low level of risk (Trabelsi & Naifar, 2017). The results of this study indicate that the NPF value has decreased following the sharia banking statistics report (2019). Most of the financing uses the Musharaka and Mudharabah contracts where the contract is at low risk with the profit-sharing scheme.  Islamic banking also needs to improve the capability of human resources to strengthen the analysis, risk mitigation, and management relationships (Azmat, Skully, & Brown, 2014). So that Islamic banking as an intermediary institution needs to carry out its functions effectively and efficiently (Rahman, Sulaiman, & Said, 2018). This can be done by paying close attention to the conditions of the prospective financing customers so that there are no problematic financing in the future (Rahman, Zheng, Ashraf, & Rahman, 2018). In addition, an effective monetary policy in considering the nature of Islamic banking is very much needed to manage the supply of credit in the economy (Rashid, Hassan, & Shah, 2020). The quality of financing and the capital ratio have an important role in financing risk (Misman, Bhatti, Lou, Samsudin, & Rahman, 2015).  (2019) which shows that Islamic banking has significantly lower credit risk. The overall development of Islamic banking has no impact on credit constraints and has a positive impact on access to credit (Léon & Weill, 2018). Financing quality with good financing potential is an attractive opportunity by utilizing diversification which is a very attractive opportunity for investors to invest (Bhuiyan, Rahman, Saiti, & Ghani, 2019). Islamic banking needs to make the composition of the financing portfolio so that it can obtain the desired returns according to market risk (Narayan, Phan, Narayan, & Bannigidadmath, 2017).

CONCLUSION
The results of the study indicate that there are NPF values, there is an element of heteroscedasticity, which means that the residual variant of the data is not constant so that it is processed using the TARCH model to get a picture of the movement of NPF values in the coming months. The tendency of forecasting done using the TARCH model decreases. This is inseparable from the current ability of Islamic banking to manage financing risks by avoiding high risk sectors.
Overall, several policy implications that can be formulated include increasing the intensity of outreach related to the development and promotion of superior loan products to attract debtors, improve the quality of human resources, provide competitive interest rates and returns along with other facilities that can be obtained by borrowers, and service improvement and providing easy access to loan applications.