THE DYNAMIC LINK BETWEEN ISLAMIC AND CONVENTIONAL DEPOSIT RATES IN A DUAL BANKING SYSTEM

Purpose — This study empirically assesses the extent to which the conventional deposit rate (CDR) affects the Islamic deposit rate (IDR) in Indonesia and Malaysia within the dual banking system. Design/Methodology/Approach — This study uses non-linear autoregressive distributed lag (NARDL) and panel cointegration. Monthly data are employed, but the time period for the two countries examined is different because of data availability. The study thus covers the period 2009:M1 to 2020:M12 for Indonesia and 2000:M1 to 2020:M12 for Malaysia. Findings — The findings confirm evidence of the long-run link between IDR and CDR, where the IDRs in Indonesia and Malaysia asymmetrically respond to changes in CDRs. In addition, Indonesia’s IDRs adjust faster in response to the decline in CDRs compared to increases in CDRs. However, Malaysia’s IDRs adapt faster in response to increases in CDRs than their decreases. The panel cointegration results reinforce the asymmetric findings. Originality/Value — To the best of our knowledge, this paper is the first study to examine the extent to which IDRs asymmetrically respond to CDRs in a dual baking system in Indonesia and Malaysia. Practical Implications — Islamic banks (IBs) follow CDRs in determining IDRs due to uncompetitive IDRs, implying that IBs suffer from displaced commercial risk. Therefore, IBs may adopt a policy to address liquidity issues through investment risk reserves (IRR) and profit equalization reserves (PER) to reduce the distinctive gap between IDRs and CDRs.


INTRODUCTION
Islamic banks (IBs) are alternative financial intermediaries with different risks from conventional banks (CBs), and they have experienced fast growth since the 1980s. However, the rapid growth of Islamic banking globally is criticised because IBs' practices are, to some extent, similar to those of CBs (Khan, 2010;Hamza, 2016). One of the criticisms relates to IBs declaring that they practise profit-and-loss sharing (PLS) financing through mushārakah and muḍārabah, nevertheless, these two types of financing combined comprise a much smaller share of total financing than murābaḥah, which applies the profit margin contract (Warninda et al., 2019). IBs' products are also based on Islamic principles, but they significantly resemble those of CBs (Chong & Liu, 2009). Moreover, IBs may peg to CBs' interest rate as a benchmark in determining their rates (Saraç & Zeren, 2015;Nechi & Smaoui, 2019).
This study explores the response of Islamic deposit rates (IDRs) to conventional deposit rates (CDRs) in Indonesia and Malaysia, which both practice the dual banking system. These countries were selected for some plausible reasons. Indonesia, for example, is the largest Muslim country and started Islamic banking in 1992. After the enactment of Law No. 23 (2008) relating to IBs, Indonesian IBs grew rapidly and performed relatively well as compared to CBs (Sutrisno & Widarjono, 2018). As for Malaysia, it implemented Islamic banking much earlier, following the passage of Islamic Banking Act 1983. This sector substantially grew in terms of the number of IBs and their market share after the Asian economic crisis in 1998 (Chong & Liu, 2009). According to the Islamic Finance Development Report 2021, Malaysia and Indonesia are ranked as the third and tenth largest countries in the global Islamic banking industry, with total assets of USD252 billion and USD39 billion in 2020, respectively (Refinitiv, 2021). However, IBs' market shares are small, with 6 per cent in Indonesia and 23 per cent in Malaysia.
This study is expected to contribute to the existing empirical literature in the following ways: 1. Previous studies applied the symmetric effect of CDRs on IDRs, such as Anuar et al. (2014) and Saraç and Zeren (2015). Contrary to the existing studies, this study employs the asymmetric effect of CDRs on IDRs to enhance the previous literature, such as Sukmana and Ibrahim (2017). 2. This study applies panel cointegration to further examine the link between IDRs and CDRs in the dual banking environment. This aims to strengthen the power of the test by applying panel data (Maddala & Wu, 1999).
This study comprises several sections. After the introduction, the second section comprises the literature review. The third section presents the methodology and data, while the fourth section delineates the findings and discussion. Finally, the last section concludes the research and discusses its implications.

REVIEW OF LITERATURE Islamic and Conventional Deposit Rates
IBs in Muslim-majority countries have been implemented after consumers had grown accustomed to conventional banking services for a long time. Furthermore, IBs, as new participants in the dual banking system, need to compete with CBs. IBs' products and services have thus had a tendency to follow the products and services of CBs, mainly because of the small number of loyal consumers who wish to avoid interest. Often, therefore, IBs' products and services are similar to those of their conventional counterparts in a dual banking environment (Khan, 2010;Azmat et al., 2015). Consequently, IB rates have to follow the interest rate, and as the IB rates deviate from interest, they have to adjust towards the long-run equilibrium interest rates, which are closely pegged (Chong & Liu, 2009). Some studies have also found that IBs' customers are very responsive to interest rate changes. As CDRs rise, customers tend to take their money back and deposit them in CBs, especially large customers (Aysan et al., 2018). Ismal (2011) also confirmed that IBs' depositors withdraw their funds from Indonesian IBs as CDRs increase. Additionally, some studies affirmed that the interest rate is negatively linked to Islamic bank deposits Kasri & Kassim, 2009;Abduh, 2015). Some existing studies using the symmetric relationship between IDRs and CDRs found that CDRs are an anchor for the IDRs in Malaysia's Islamic banking industry (Chong & Liu, 2009;Zainol & Kasim, 2010;Anuar et al., 2014;Saeed et al., 2021). Numerous studies also documented that the IDRs mimic the CDRs in Turkey (Ergeç & Arslan, 2013;Ergec & Kaytanci, 2014;Cevik & Charap, 2015;Saraç & Zeren, 2015). Kasri and Kassim (2009) further found that the CDR significantly affects the IDR in Indonesia. Sukmana and Ibrahim (2017) documented that IDRs asymmetrically follow the rise and fall in CDRs in the Malaysian Islamic banking industry, where the IDRs adjust at a slower speed to falling CDRs. Mushafiq and Sehar (2021) found that IB deposit rates are strongly influenced by CB deposit rates in the short run for Pakistani banks.
However, some previous empirical studies found opposite results whereby IDRs are not linked to conventional interest rates. Yusof et al. (2015) assessed the relationship between IDRs and CDRs in the Gulf Cooperation Council (GCC) countries. They found that there is no relationship between IDRs and CDRs in the long run. Another study showed that the interaction between IDRs and CDRs is weak for 20 major banks located in some emerging and developed countries (Jawadi et al., 2016). By investigating the causal link between IDRs and CDRs in Turkey, Yuksel (2017) found that IBs do not determine their IDRs by pegging to CDRs. Furthermore, CBs set interest rates without any reference to the IDR.

Islamic Interbank and Conventional Interbank Rate
The Islamic interbank rate (IIR) also links with the conventional interbank rate (CIR). Strong comovement between IIR and CIR has been found to prevail in the Malaysian money market (Bacha, 2008;Ito, 2013) and in five GCC countries (Nechi & Smaoui, 2019). Moreover, in the global money market, there is evidence of the correlation between the Islamic interbank benchmark rate (IIBR) and the London interbank offer rate (LIBOR) for every maturity (Azmat et al., 2020). In contrast, Tlemsani (2020) documented a strong negative correlation between the IIBR and LIBOR.
Initially, some studies examined the symmetric link between IDRs and CDRs (Chong & Liu, 2009;Saraç & Zeren, 2015). However, consistent with the banking literature, the possibility of an asymmetrical link between deposit rates and lending rates arises as a result of some factors, such as imperfect markets, collusion behaviour and consumer responses to changes in the interest rate (Nguyen & Islam, 2010;Holmes et al., 2015). The premise of an imperfect market claims that banks may capitalise on their market power through monopoly behaviour or collusion to benefit from upward changes in lending rates which adjust faster than upward changes in deposit rates. Meanwhile, consumers tend to refuse high lending rates and low deposit rates. Therefore, deposit rates are likely to be sticky downwards while lending rates are naturally sticky upwards.
Accordingly, the link between the IDR and the CDR may be asymmetric (Sukmana & Ibrahim, 2017). Therefore, this study can express the long-run asymmetric relationship between those rates in the following regression equation: where is IDR and is CDR. Variables !&# $ and !&# & indicate an increase and reduction in ! , showing the asymmetric response of the IDR to a rise and fall in the CDR. Several conditions may arise from the asymmetric relationship between the IDR and the CDR. Firstly, the consumer determines the behaviour of IBs in line with the consumer reaction theory ( # > % ). Secondly, the IB establishes its rate following the market concentration theory ( # < % ). Thirdly, the Islamic rate does not follow the interest rate.
The non-linear autoregressive distributed lag (NARDL) is applied to examine the asymmetric effect of CDRs on IDRs because the symmetric model generates a biased estimation that ignores such asymmetric relationships (Shin et al., 2014). As the asymmetric effect occurs, the IDRs respond differently to changes in the CDRs depending on the level of increase or decrease in the CDRs.
Variables !&# $ and !&# & are partial sums of a rise and fall in ! . They are computed as follows: The NARDL model of equation (1) is as follows: From equation (4), then, the long-run asymmetric effects of an increase and decrease in CDRs on IDRs are computed as , respectively. The short-run asymmetric impacts of a rise and fall in CDRs on IDRs are computed as # = ∑ & , respectively. Some steps are applied to estimate the NARDL model, as follows: 1. This study applies the ordinary least squares (OLS) method and uses the general-to-specific method by sequentially dropping insignificant lags to arrive at the final model. 2. It tests the long-run link between CDRs and IDRs using two cointegration tests. The first cointegration method tests the null hypothesis # = 0 following 012 statistic test (Banerjee et al., 1998). Meanwhile, the second cointegration method checks the null hypothesis # = % = 0 following the 344 statistic test (Pesaran et al., 2001). 3. When cointegration exists, then this paper assesses the asymmetric response of IDRs to CDRs. The null hypotheses of the short-run and long-run asymmetric response of IDRs to CDRs are # = % and # = % , respectively. 4. This study calculates the asymmetric dynamic multiplier effect for percentage change in the conventional deposit rate (∆ !&# $ , ∆ !&# & ) on IDRs by applying the formula provided by Shin et al. (2014):

Panel Cointegration
Due to similar Islamic banking practices in Indonesia and Malaysia, this study employs the panel method to examine the response of IDRs to CDRs in both countries. The panel cointegration method is applied to further explore the effect of CDRs on IDRs. Some steps are carried out to estimate the panel cointegration, as follows: 1. This study tests the stationarity of panel variables by following Levin et al. (2002) and Im et al. (2003). 2. Subsequently, after the integration order of the variables is established, the existence of cointegration between the variables is examined using Pedroni (1999). 3. This study then assesses the impact of CDR on IDR using Fully-modified OLS (FMOLS) and Dynamic OLS (DOLS).

Data
The data are obtained from the Indonesian Financial Services Authority and the Central Bank of Malaysia, and the IDRs and CDRs are associated with each maturity. The time deposit rates of IBs and CBs encompass 1-, 3-, 6-and 12-month maturities. This study uses monthly data, but the time is different across the two countries due to the data availability, covering the periods 2009:M1 to 2020:M12 and 2000:M1 to 2020:M12 for Indonesia and Malaysia, respectively. Table 1 shows the descriptive statistics as a preliminary analysis of the data. Several interesting findings emerge from these descriptive statistics. Overall, the average CDRs are found to be higher than the IDRs in both countries. However, the standard deviation of the CDRs is higher than the IDRs. These findings clearly indicate that the IDR is less variable than the CDR. Also, the correlations of each maturity between IDRs and CDRs are relatively high. These findings support some previous literature on Malaysian IBs (Chong & Liu, 2009;Sukmana & Ibrahim, 2017

NARDL Results: Indonesian Islamic Deposit Rates
Before estimating NARDL, this study checks the stationary data to guarantee that they fit the model. The Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests are employed to decide on the variable stationarity. Table 2 reveals the findings, which report that several variables are stationary while some are not. Nevertheless, the first difference data are stationary for all variables, implying that the NARDL is an appropriate model. This study focused on the NARDL results to examine the asymmetric response of IDRs to CDRs. Table 3 reveals the NARDL findings of the Indonesian IDRs with optimal lag order up to 12. This study reports the results and diagnostic tests at the top of the table. The test of cointegration is presented in the middle part of the table, employing both 012 and 344 statistic tests. All null hypotheses for all maturity rates are rejected, meaning that that the long-run link between IDRs and CDRs exists. The next row exhibits the long-run asymmetric test using the Ftest statistic. The null hypothesis for all maturities is rejected. These findings conclude that the IDRs asymmetrically respond to changes in the corresponding CDRs.
The long-run asymmetric coefficients of positive CDR ( $ ) and negative CDR ( & ) are presented in the bottom part of Table 3. All coefficients of $ and & are significant at 1 per cent. For example, the long-run asymmetric coefficients of $ and & for the 1-month maturity rate are 0.801 and 0.901, respectively. These findings show that a 1 per cent rise in CDR increases IDR by 0.801 per cent. Meanwhile, a fall in the CDR by 1 per cent is associated with a fall in the IDR by 0.901 per cent. Overall, the asymmetric pricing behaviour of the IDRs grows weaker for longer maturities, but the difference is wider. Interestingly, the CDR pass-through to the IDR is larger for & as compared to $ for all maturities. The results may suggest that the IB establishes its rate due to market concentration. The Indonesian Islamic banking market is imperfect and close to the tight oligopoly market with concentration ratio of the four largest IBs (CR-4) of 68.86 per cent and 49.31 per cent in 2010 and 2020, respectively (Widarjono et al., 2020a).
Lastly, this study exhibits asymmetric cumulative dynamic multipliers for both deposit and financing rates. The Indonesian Islamic-conventional deposit rate dynamic multiplier over the 80month horizon, with a 90 per cent confidence interval for the different maturities (i.e., 1-month, 3month, 6-month and 12-month), is reported in Figure 1. These graphs indicate that IDRs respond to CDRs for all maturities. A rise (fall) in the IDRs likely responds to an increase (fall) in CDRs.

NARDL Results: Malaysian Islamic Deposit Rates
The results of NARDL for the Malaysian IDRs with optimal lag order up to 12 are exhibited in Table 4. According to 012 and 344 statistics, this study rejects the null hypothesis for all maturity periods, implying that a long-run link between IDRs and CDRs exists. The null hypothesis of no long-run asymmetric effect for all maturities is rejected. The results show the asymmetric effect of the CDRs on IDRs for all maturities. The null hypothesis of no short-run asymmetric effect is also rejected, except for the 1-month maturity rate. All coefficients of $ and & are significant, for instance, the coefficients of $ and & are 0.850 and 0.660 for the 1-month maturity rate, respectively. These results suggest that a 1 per cent rise in the CDR generates roughly 0.850 per cent increase in the IDR. Also, a 1 per cent reduction in the CDR results in approximately 0.600 per cent decrease in the IDR. The asymmetric pricing behaviour of the IDR tends to be stronger for longer maturities, and the CDR pass-through to the IDR is larger for $ than for & for all maturities. The results may imply that the IDRs in Malaysia are related to the consumer reaction hypothesis, confirming the previous study by Sukmana and Ibrahim (2017). The Malaysian Islamic-conventional deposit rate dynamic multipliers over the 80-month horizon, with a 90 per cent confidence interval for all maturity periods (i.e., 1-month, 3-month, 6-month and 12-month), are presented in Figure 2. The figure indicates that IDRs asymmetrically respond to the CDRs for all maturities. Therefore, a rise (fall) in the IDRs is associated with a rise (fall) in the CDRs.
The results from the DOLS model imply that a 1 per cent rise in the 1-month CDR leads to a 0.647 per cent rise in IDR. Also, a 1 per cent decrease in the 1-month CDR leads to a 0.749 per cent reduction in IDR. The findings from the FMOLS model suggest that a 0.616 per cent increase in the 1-month IDR is related to a 1 per cent increase in CDR. Furthermore, a 0.704 per cent increase in the 1-month IDR is linked to a 1 per cent increase in CDR. In general, the results from the panel cointegration method reinforce these findings from the country-by-country NARDL method. 940 Notes: *** is significant at 1%, ** is significant at 5% and * is significant at 10%, respectively. The Akaike Info Criterion (AIC) method is applied to select the number of leads and lags. Source: Authors' own

DISCUSSION
These results document that the IDRs strictly follow the CDRs, confirming the existing empirical studies such as Chong andLiu (2009), Cevik andCharap (2015), and Sukmana and Ibrahim (2017). As IBs could not attain the yield offered in the PLS rates, IBs may peg their PLS rates to interest rates (Hamza, 2016). Furthermore, IBs may have less experience in monitoring equity-financed projects based on PLS contracts that generate high financing risk due to the agency problem (Mushafiq & Sehar, 2021;Al-Harbi, 2022). Accordingly, they prefer the use of the murābaḥah (mark-up) contract, which is the most common form of IB financing (Widarjono et al., 2022). This result appears to be in line with the consumer theory of Islamic deposit products (Sukmana & Ibrahim, 2017). Within a dual banking system, IB depositors have two choices in investing their funds to yield the highest return. Indeed, religiosity is not the main factor for IB depositors to save their money in IBs in a dual banking system (Azmat et al., 2020;Utomo et al., 2021). Consequently, IBs need to follow CDRs in determining the IDRs, especially as interest rates rise. When this is not done, IB depositors will tend to withdraw their funds from IBs.
Numerous studies found that internal and external factors affect customers' withdrawal of funds from IBs. The internal factors result from uncompetitive IDRs when CBs give higher rates for any maturity (Ismal, 2011). Accordingly, several studies reported that a negative link exists between IDRs and CDRs, which is known as displaced commercial risk (Kasri & Kassim, 2009;Kassim et al., 2009;Abduh, 2015). Therefore, to withstand displaced commercial risk, IBs need to remain competitive and counter commercial pressures to prevent customers from switching to their conventional competitors.
The displaced commercial risk can be reduced by applying the investment risk reserve (IRR) and profit equalization reserve (PER) policy (Toumi et al., 2019;Touri et al., 2020). The IRR is a portion of an investor's income that is appropriated by the IB, after allocating the muḍarīb's share, to cover for future losses on investments financed by muḍārabah, mushārakah and murābaḥah contracts. Meanwhile, PER is a portion of the gross income from the murābaḥah contract that is appropriated by the IB before its allocation to the depositor to provide a more equitable return to the muḍarīb. Both policies obviously increase the return of Islamic deposits and lower the distinct gap between IDRs and CDRs.