Date of submission: May 31, 2021; date of acceptance: August 5, 2021. * Contact information: mdaliashrafaiba@gmail.com, Army Institute of Business Ad- ministration, Jalalabad Cantonment, Sylhet-3104, Bangladesh, phone: +8802996642777; ORCID ID: https://orcid.org/0000-0003-2146-6267. Copernican Journal of Finance & Accounting e-ISSN 2300-3065 p-ISSN 2300-12402021, volume 10, issue 4 Ashraf, M. A. (2021). The Impact of Mobile Financial Services on the Usage Dimension of Finan- cial Inclusion: an Empirical Study from Bangladesh. Copernican Journal of Finance & Accounting, 10(4), 9–25. http://dx.doi.org/10.12775/CJFA.2021.012 Md ali ashraf* Army Institute of Business Administration the iMpact of Mobile financial services on the usage diMension of financial inclusion: an eMpirical study froM bangladesh Keywords: MFS, financial inclusion, COVID-19. J E L Classification: G20, G21, G23, E42, O33. Abstract: A plethora of studies have investigated how Mobile Financial Services (MFS) induces financial inclusion around the world. However, research in the context of Bang- ladesh was rather limited. Hence, the primary objective of this paper was to investigate whether there was a statistically significant relationship between MFS and financial in- clusion, measured by two time series variables – the number of MFS agents and number of registered MFS users per 100,000 of population, from 2017 to 2020. For analyzing the relationship between these two variables, multiple statistical methods were employed – including Vector Auto Regression, Cointegration and Granger Causality. The analysis revealed that both time series variables had an increasing trend with time. More im- portantly, the analysis specified that there was no statistically significant relationship between MFS, measured by the number of agents and the ‘usage’ dimension of financial inclusion, measured by number of registered MFS users in the context of Bangladesh. Moreover, the study was unable to find any significant changes in the trends of these variables that could be attributed to the COVID-19 pandemic in Bangladesh. Mohammad Ali Ashraf10  Introduction Technological inventions are disrupting the way businesses conduct their ac- tivities and, at the same time, transforming the means through which individ- uals engage in financial transactions. Information technology, combined with innovative communication systems and tools are bringing enormous changes throughout the globe. One of the major changes is occurring in the method peo- ple use for completing their financial activities. According to a study by World Bank (2012), only half of the global population maintained accounts with a for- mal financial institution. One of the commonly reported barriers of accessing fi- nancial services was the physical distance of service facilities (Demirguc-Kunt & Klapper, 2012). This is where mobile financial services come into considera- tion. By using mobile devices and networks, a huge number of unbanked indi- viduals, especially low-income and rural consumers around the globe, are get- ting access to previously unavailable financial products and services (Kanobe, Alexander & Bwalya, 2017). To understand how mobile devices are making financial services available to a significant portion of the global population, it is essential to understand two notions – Mobile Financial Services (MFS) and Financial Inclusion. Mobile Financial Services (MFS) generally refers to the use of mobile or handheld communication devices to access banking services. Bangladesh Bank, the central bank of Bangladesh defines MFS as – “an approach to offering finan- cial and banking services via mobile wireless networks which enables for user to execute banking transactions.” (Bangladesh Bank, 2012). The idea of financial inclusion is to enable mass people to engage in finan- cial activities through official channels. Sarma and Pais (2011) defined finan- cial inclusion as a process that ensures the ease of access, availability and us- age of the formal financial system for all members of an economy, whereas Dev (2006) defined financial inclusion as the delivery of banking services at an af- fordable cost for a massive section of underprivileged members of an economy. A substantial number of scholarly studies demonstrated that mobile finan- cial services played a vital role in facilitating financial inclusion around the world, specially in underdeveloped and developing countries. This topic has been discussed further in the literature review section. Likewise, to under- stand how MFS is facilitating financial inclusion, it is imperative to look at some of the indicators that measures the extent of financial inclusion. the imPAct of mobile finAnciAl services… 11 Indicators of financial inclusion can measure three dimensions of financial inclusion including ‘Usage’, ‘Access’ and ‘Quality’ (Pearce & Ortega, 2012). Ac- cess indicators show the depth of outreach of financial services such as number of bank branches or number of ATMs per 100,000 of the population, whereas usage indicators ref lect how clients use financial services by looking at data such as number of adults with an account at a formal financial institution (Sar- ma, 2008; World Bank, 2015). As this paper will attempt to investigate whether MFS is enabling financial inclusion, a similar access indicator used in this paper will be the number of MFS agents. In terms of usage dimension, the selected in- dicator will be number of registered MFS users per 100,000 of population. In addition, according to a report by Business Finance for the Poor in Bangladesh (2021), a key policy initiative planned by the authorities of Bangladesh is to in- crease availability of MFS, or in other words, to increase the number of access points or MFS agents to get more people under a formal financial system. It may be noted at this point that, research on how MFS is contributing to financial in- clusion in Bangladesh is rather limited. Hence, one of the objectives of this pa- per is to investigate whether a particular dimension of MFS, namely the num- ber of agents that provide mobile financial services, can affect the number of users of such services – as indicated by the number of registered MFS users in Bangladesh. For this purpose, four years of monthly time series data starting from January, 2017 of these two variables were used. An additional objective of this paper is to have a deeper look at the two vari- ables mentioned earlier and their behavior over the same time period in an at- tempt to discover how the pandemic affected the trends of these variables. By concentrating on these two objectives, the paper not only aims to pro- vide valuable insights for policy makers and entities that are interested in en- hancing financial inclusion, but also helps in understanding the trend of MFS and financial inclusion in Bangladesh, and if the COVID-19 pandemic had an im- pact on the trend. Literature Review Ensuring access to financial services for people from every socio-economic class of a nation is imperative for economic growth. Indeed, it has been shown that financial inclusion has positive effect on the economic growth (Iqbal & Sami, 2017; Kim, Yu & Hasan, 2018; Sharma, 2016). Moreover, financial in- Mohammad Ali Ashraf12 clusion contributes to the development of society and lessens the gap between people from different financial tiers. In addition, as more people start availing financial services, their transactions boost money f low in the economy (Da- modaran, 2013). It has also been argued that the benefits of financial inclusion reach beyond its users. Ozili (2018) stated that MFS providers and government, both benefit from a population that is financially inclusive. People that remain excluded from financial inclusion suffer from poverty and economic instability. Those who are unable to access financial services, face multi-dimensional social exclusion as well (Devlin, 2005). It has also been shown that people who are unable to access financial services incur additional expenses in managing household resources. Besides, a population facing finan- cial exclusion often have difficulties in getting employment (Corr, 2006). Addi- tionally, businesses often fail to reap the benefits of MFS as it incurs high trans- action costs (Kabir, Sadrul Huda & Faruq, 2020). As discussed earlier, financial inclusion is crucial for economic develop- ment, and it is essential to look at how mobile devices are contributing towards financial inclusion. Smartphones are becoming a household item for a signifi- cant share of the global population. Along with relatively cheap mobile internet services, smartphone users, particularly individuals fighting with poverty or those from remote areas, are availing banking and financial services that were previously inaccessible. A significant number of studies have shown evidence in favor of this phenomenon. Huge investments have been made in infrastruc- tures and applications that use mobile phones to deliver financial services, in order to bring the poor and unbanked population under the umbrella of finan- cial inclusion (Porteous, 2006; Vodafone, 2007; Bangens & Soderberg, 2008; World Bank, 2012). Likewise, Horne (1985), Frame and White (2004) found that one form of financial innovation is mobile technology, and such innova- tions have multi-dimensional effects. Innovative mobile technologies improve efficiency and effectiveness of the financial system and thus, improves financial inclusion. Additionally, one study indicated that MFS is responsible for increas- ing the volume and frequency of remittances received by users, and a small in- crease in savings by such users (Alampay, Moshi, Ghosh, Peralta & Harshanti, 2017). Kim, Zoo, Lee and Kang (2018), Mbogo (2010) also argued that delivery of financial services by means of mobile devices enables financial inclusion. Nevertheless, even in this era of information and communication technolo- gy, a huge portion of the global population remains unbanked. At the beginning of 21st century, half of the global population had no access to banking services the imPAct of mobile finAnciAl services… 13 (Chaia, Dalal, Goland, Gonzalez, Morduch & Schif, 2009). There was no signifi- cant change in this parameter as indicated by World Bank (2012). Although a newer survey by World Bank (2018) reported that the number of financially included individuals are increasing, about one-third of the population still re- mains unbanked. The situation is not very different in the case of Bangladesh. According to World Bank, in 2011, only 31.7% of people in Bangladesh had ac- cess to financial services. In 2014, this number declined slightly to 31%, and fi- nally reached 50% in 2018. In contrast, almost 70% people of South Asia had an account that provided financial services in 2018 (World Bank, 2018). Hence, it can be said that there is a huge potential to advance financial inclusion through mobile banking in Bangladesh. As mentioned earlier, policy makers of Bangla- desh are emphasizing on creating more access points to integrate more peo- ple in a formal financial system. However, a review by Duncombe and Boateng (2009) asked for additional empirical investigation to verify such causal link. Furthermore, a study conducted in Bangladesh by Siddik, Sun, Yanjuan and Ka- biraj (2014) pointed out that MFS providers should be more concerned about the factors that affect behavioral intention of people in availing and continu- ously using such services. In order to get more people under a formal financial system, the suggested policy according to the study was to enhance customer satisfaction through improving and delivering financial services via mobile de- vices. These studies lay the foundation of the main objective of the study – to check if increasing the number of MFS agents’ results in increased number of MFS users. Methodology The first task was to examine the trends of the two time series variables perti- nent to MFS usage and financial inclusion, by using monthly data from January, 2017 to December, 2020, obtained from Bangladesh Bank MFS dataset (Bang- ladesh Bank, 2021). These datasets had time series data on variables such as – number of MFS agents, number of registered and number of active MFS us- ers, total number of transactions, the volume of average daily transactions, and product wise information of MFS transactions. To address the objective of this study, only two of the time series variables with 48 observations were selected, the first one being the total number of MFS agents and the second one being the total number of registered MFS users. Mohammad Ali Ashraf14 Next, it was checked whether there was any statistically significant relation between the time series variables – total number of MFS agents and total num- ber of registered MFS users. Multiple statistical methods including Vector Auto Regression, Cointegration and Granger Causality for analyzing relationship of time series data were used in the process. The statistical application EViews was used for conducting all the analyses. Stationarity At the beginning of the analysis, data was checked for stationarity through vis- ual inspection and the Augmented Dickey–Fuller (ADF) test. ADF test is wide- ly used to check for unit roots in time series variables. A preliminary visual inspection revealed increasing trends for both variables. Then, the ADF test checked for unit root in the time series, which was conducted at 90, 95 and 99 percent confidence level on both time series variables at level and at first dif- ference. Lag length for the test was automatically selected based on Schwarz information criterion (SIC). It is imperative to note that both time series had unit roots and were non-stationary with constant, with constant and trend, and without constant & trend, and they became stationary at first difference according to the ADF test (table 1). In other words, the variables were non-sta- tionary at level and integrated of same order - I(1). Cointegration Cointegrating variables are said to have long run equilibrium relationship, meaning they will not drift apart if there is disturbance. In empirical litera- ture, cointegration between variables is frequently checked by using the Jo- hansen Cointegration test. Hence, cointegration between the variables under study was checked by using Johansen Cointegration test (Johansen, 1988). Be- fore checking for cointegration, an initial Vector Autoregressive (VAR) model was estimated with both time series at level. It may be mentioned here that VAR is a statistical model to analyze relationship between time series variables. Af- ter estimating the initial VAR, tests were conducted to select an appropriate lag length for checking cointegration. Johansen Cointegration test was then conducted based on five different sets of assumptions for allowing “No deterministic trend”, “Linear deterministic the imPAct of mobile finAnciAl services… 15 trend” and “Quadratic deterministic trend” in the cointegrating equation. The test revealed that there was no cointegration between the two time series vari- ables. Table 2 provides a summary of the cointegration test. Vector Autoregressive Model Vector autoregression (VAR) is a statistical model that is frequently used to ex- amine the relationship between multiple time series variables. Impulse response functions were then projected in the context of the devel- oped VAR model by using Cholesky Decomposition method with dof adjusted. Since the main objective of the paper was to investigate whether number of MFS agents can inf luence the number of MFS users, number of MFS agents was first in Cholesky ordering followed by number of registered MFS users. Then, impulse response graphs were analyzed to check how changes in number of MFS agents impact the number of registered MFS users. In addition to the impulse response function, forecast error variance de- composition was used to investigate how the change in number of MFS agents explains variation in number of registered MFS users. Granger Causality Test Additionally, Granger causality test was used to check whether any of the vari- able Granger caused the other variable. Granger causality test is a statistical hypothesis test that examines if one time series can predict the variability in another time series (Granger, 1969). The parameters for this test were based on the optimal VAR model estimated during the previous step. Results & Discussion During the time period of 2017 to 2020, both number of MFS agents and num- ber of registered MFS users were increasing steadily. In January 2017, the number of MFS agents was merely 723,112; which increased by more than 46% to 1,058,897 by December of 2020. This increasing trend is clearly vis- ible in chart 1. During the same time period, number of registered MFS users per 100,000 individuals increased to 993 from 419, an increase of more than 136% (chart 2). Mohammad Ali Ashraf16 Chart 1. Number of MFS agents registered MFS users per 100,000 individuals increased to 993 from 419, an increase of more than 136% (chart 2). Chart 1. Number of MFS agents Source: own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). From the charts, it is also visible that both time series shows an increasing trend, but no apparent seasonality was revealed, nor was there any indication of cyclic behavior. From visual inspection of the charts, it can also be stated that both time series were non stationary. Chart 2. Number of register ed MFS users per 100,000 of populati on 600,000 700,000 800,000 900,000 1,000,000 1,100,000 2017 2018 2019 2020 No. of MFS agents 400 500 600 700 800 900 1,000 2017 2018 2019 2020 No. of registered clients per 100,000 of population S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). From the charts, it is also visible that both time series shows an increasing trend, but no apparent seasonality was revealed, nor was there any indication of cyclic behavior. From visual inspection of the charts, it can also be stated that both time series were non stationary. Chart 2. Number of registered MFS users per 100,000 of population registered MFS users per 100,000 individuals increased to 993 from 419, an increase of more than 136% (chart 2). Chart 1. Number of MFS agents Source: own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). From the charts, it is also visible that both time series shows an increasing trend, but no apparent seasonality was revealed, nor was there any indication of cyclic behavior. From visual inspection of the charts, it can also be stated that both time series were non stationary. Chart 2. Number of register ed MFS users per 100,000 of populati on 600,000 700,000 800,000 900,000 1,000,000 1,100,000 2017 2018 2019 2020 No. of MFS agents 400 500 600 700 800 900 1,000 2017 2018 2019 2020 No. of registered clients per 100,000 of population S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). the imPAct of mobile finAnciAl services… 17 Impact of COVID-19 Like all other countries in the world, Bangladesh also faced the consequence of the COVID-19 pandemic. On March 8, 2020, the first case of COVID-19 was de- tected in Bangladesh, and the government initiated a countrywide shutdown on 26 March, 2020. As movement of individuals was restricted, commercial ac- tivities were suspended and formal financial service organizations like banks remained inactive, it was assumed that MFS activities will see a substantial rise (Islam, Talukder, Siddiqui & Islam, 2020). Although there were changes in the trends of these variables after the month of March, according to data 6 month prior and after March 2020, the evidence was inconclusive to attrib- ute such changes to the pandemic. During the 48 months analyzed in the study, number of agents had an average growth rate of 0.82% with a standard devia- tion of 0.92%, while number of registered users had an average growth rate of 1.89% with a standard deviation of 2.78%. In fact, the monthly average growth rate of the number of agents was 0.60% during the six months prior to the first case detection, whereas this average growth rate dropped down to 0.40% dur- ing the six months afterwards. In case of number of registered users, during the six months prior to pandemic, the growth in registered users were 1.79%, and 2.33% percent during the six months after march. When compared to the standard deviations, it can be stated that there were no abnormal changes in the growth rate after March 2020. Stationarity As discussed earlier, visual inspection of both time series revealed that both exhibit increasing trend and are non-stationary. This finding was also support- ed by the ADF test of stationarity, results of which are presented in table 1. Both variables had a unit root and thus were non-stationary at level. The ADF test also showed that at 95% confidence level, both variables became station- ary at first difference. Mohammad Ali Ashraf18 Table 1. Unit Root Test Results Table (ADF) Null Hypothesis: the variable has a unit root At Level NO_ _OF_REGISTERED_ CLIENTS_PER_100000 NO_ _OF_MFS_AGENTS With Constant t-Statistic 0.0144 -1.2596 Prob. 0.9550 0.6403 no no With Constant & Trend t-Statistic -1.2911 -2.6531 Prob. 0.8781 0.2601 no no Without Constant & Trend t-Statistic 6.1373 6.9115 Prob. 1.0000 1.0000 no no At First Difference d(NO_ _OF_REGISTERED_ CLIENTS_PER_100000) d(NO_ _OF_MFS_AGENTS) With Constant t-Statistic -4.4037 -9.4196 Prob. 0.0010 0.0000 *** *** With Constant & Trend t-Statistic -10.8314 -9.3775 Prob. 0.0000 0.0000 *** *** Without Constant & Trend t-Statistic -2.3026 -1.5496 Prob. 0.0221 0.1127 ** no Notes: a: (*) Significant at the 10%; (**) Significant at the 5%; (***) Significant at the 1% and (no) Not Sig- nificant b: Based on SIC c: Probability based on MacKinnon (1996) one-sided p-values. S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). the imPAct of mobile finAnciAl services… 19 Co- integration To check for cointegration between the two time series variables, Johansen Cointegration test was used in this study. The results of the test (table 2) in- dicated that at the selected lag length of 2 specified by AIC and FPE, there was no cointegration. Both Trace and Maximum Eigenvalue tests showed similar results indicating the absence of any cointegration among the two time series variables under investigation. Table 2. Summary of Cointegration test Date: 03/17/21 Time: 13:59 Sample: 2017M01 2020M12 Included observations: 45 Series: NO_ _OF_MFS_AGENTS NO_ _OF_REGISTERED_CLIENTS_PER_100000 Lags interval: 1 to 2 Selected (0.05 level*) Number of Cointegrating Relations by Model Data Trend: None None Linear Linear Quadratic Test Type No Intercept, Intercept, Intercept, Intercept, Intercept, No Trend No Trend No Trend Trend Trend Trace 1 0 0 0 0 Max-Eig 1 0 0 0 0 * Critical values based on MacKinnon-Haug-Michelis (1999). S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). Vector Autoregressive Model Impulse Response Analysis of VAR model Impulse response analysis investigates how a time series variable responds to a shock in another time series variable. Based on the VAR model developed in the previous step, four impulse response graphs were produced. Since the ob- jective of the study was to check how a change in number of MFS agents chang- es the number of registered MFS users, among the four graphs, only the one that depicts the response of number of registered users to a shock in number of agents is highlighted here (chart 3). Mohammad Ali Ashraf20 Chart 3. Impulse response: response of number of registered MFS users to a change in number of MFS agents impulse response graphs were produced. Since the objective of the study was to check how a change in number of MFS agents changes the number of registered MFS users, among the four graphs, only the one that depicts the response of number of registered users to a shock in number of agents is highlighted here (chart 3). Chart 3. Impulse response: response of number of registered MFS users to a change in number of MFS agents Source: own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). From the graph, it is evident that initially there is a negative impact on the number of registered MFS users, as indicated by the line starting below 0, when there is a shock in the number of MFS agents. Eventually, this negative impact fades away as the line moves from below 0 to above 0 and then merges into 0. One way to interpret this finding is that, when the number of MFS agents goes up by 1 unit, during the next two months, the number of registered MFS users per 100,000 of population goes down by more than 2 units. During the next 6 months, number of registered MFS users is marginally positively affected – after which, this effect starts to decay, eventually becoming zero, for 1 unit of positive change in number of MFS agents. -2 0 2 4 6 8 1 2 3 4 5 6 7 8 9 10 11 12 Response of D(NO__OF_REGISTERED_CLIENTS_PER_100000) to D(NO__OF_MFS_AGENTS) S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). From the graph, it is evident that initially there is a negative impact on the num- ber of registered MFS users, as indicated by the line starting below 0, when there is a shock in the number of MFS agents. Eventually, this negative impact fades away as the line moves from below 0 to above 0 and then merges into 0. One way to interpret this finding is that, when the number of MFS agents goes up by 1 unit, during the next two months, the number of registered MFS users per 100,000 of population goes down by more than 2 units. During the next 6 months, number of registered MFS users is marginally positively affected – after which, this effect starts to decay, eventually becoming zero, for 1 unit of positive change in number of MFS agents. Forecast Error Variance Decomposition of VAR Model In a similar manner to Impulse Response analysis, forecast error variance de- composition (FEVD) shows how much of the variability of one time series is caused by shocks in another time series. By looking at the variance decomposi- tion of number of registered clients in table 3, it can be stated that regardless of the imPAct of mobile finAnciAl services… 21 short or long term, number of MFS agents does not significantly affect the num- ber of registered MFS users. For instance, during the 6th period, only about 16% of the forecast error variance in number of registered users was due to number of MFS agents, which was deemed insignificant. Table 3. Forecast Error Variance Decomposition Variance Decomposition of D(NO_ _OF_MFS_AGENTS): Period S.E. D(NO_ _OF_MFS_AGENTS) D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) 1 4342.483 100.0000 0.000000 2 4589.611 99.45571 0.544294 3 4715.778 96.00657 3.993425 4 4732.140 95.80261 4.197390 5 4755.958 95.08286 4.917138 6 4760.689 94.99087 5.009126 7 4765.463 94.85346 5.146536 8 4766.764 94.82494 5.175063 9 4767.772 94.79699 5.203013 10 4768.110 94.78912 5.210883 11 4768.331 94.78316 5.216845 12 4768.415 94.78112 5.218883 Variance Decomposition of D(NO_ _OF_REGISTERED_CLIENTS_PER_100000): Period S.E. D(NO_ _OF_MFS_AGENTS) D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) 1 8.894830 16.98720 83.01280 2 8.938255 17.12442 82.87558 3 9.131167 16.67070 83.32930 4 9.140166 16.79548 83.20452 5 9.169914 16.80620 83.19380 6 9.173320 16.83979 83.16021 7 9.178825 16.84695 83.15305 8 9.179872 16.85441 83.14559 9 9.180970 16.85662 83.14338 Mohammad Ali Ashraf22 Variance Decomposition of D(NO_ _OF_REGISTERED_CLIENTS_PER_100000): Period S.E. D(NO_ _OF_MFS_AGENTS) D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) 10 9.181261 16.85828 83.14172 11 9.181492 16.85886 83.14114 12 9.181568 16.85924 83.14076 Cholesky Ordering: D(NO_ _OF_MFS_AGENTS) D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). Granger Causality Test Granger causality test between the two variables was conducted with the same parameters of the VAR model discussed in the previous section. From the re- sults of the test (table 4), it was concluded there was no Granger causality among the variables, as indicated by a p value of higher than 0.05. It may also be mentioned here that none of the variable Granger caused the other variable. Table 4. VAR Granger Causality/Block Exogeneity Wald Tests Date: 03/18/21 Time: 20:13 Sample: 2017M01 2020M12 Included observations: 45 Dependent variable: D(NO_ _OF_MFS_AGENTS) Excluded Chi-sq df Prob. D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) 3.127916 2 0.2093 All 3.127916 2 0.2093 Dependent variable: D(NO_ _OF_REGISTERED_CLIENTS_PER_100000) Excluded Chi-sq df Prob. D(NO_ _OF_MFS_AGENTS) 1.243859 2 0.5369 All 1.243859 2 0.5369 S o u r c e : own study based on Bangladesh Bank MFS dataset (Bangladesh Bank, 2021). Table 3. Forecast… the imPAct of mobile finAnciAl services… 23  Conclusion During the period of 2017 to 2020, Bangladesh realized a positive growth in the number of MFS agents and in the number of registered MFS users, and the COVID-19 pandemic that started in March of 2020 did not cause any significant change in the increasing trend of these two variables. Although both the number of agents and number of users were steadily in- creasing simultaneously, however, in this particular case correlation did not lead to causation. The study was unable to find and long-term cointegrating relationship between these two variables, nor was it able to find Granger cau- sality. Moreover, Impulse response and variance error decomposition methods in the VAR context revealed that increasing the number of MFS agents initially had a negative impact on the number of MFS users, followed by a marginal posi- tive impact for a couple of months. Ultimately, any effect of increasing the num- ber of MFS agents decayed out rather rapidly. Hence, it was concluded that increasing access to mobile financial services may not necessarily lead to more people being financially inclusive. The policy makers should refrain from solely relying on increasing the number of MFS ac- cess points or agents. What could be a better option to improve financial inclusion in Bangladesh? The answer to this question requires additional investigation. One potential option based on earlier research is that, rather than increasing the number of MFS access points at this point, it would be relatively more beneficial to focus on increasing customer satisfaction by improving the services offered by MFS providers.  References Alampay, E.A., Moshi, G.C., Ghosh, I., Peralta, M.L., & Harshanti, J. (2017). The impact of mobile financial services in low- and lower middle-income countries. Ontario: Inter- national Development Research Centre. Bangens, L., & Soderberg, B. (2008). Mobile banking: financial services for the unbanked? Stockholm: The Swedish Programme for ICT in Developing Regions. Bangladesh Bank (2012). Mobile financial services in Bangladesh: an overview of Mar- ket Development, https://www.bb.org.bd/pub/research/policypaper/pp072012. pdf (accessed: 11.01.2021). Mohammad Ali Ashraf24 Bangladesh Bank (2021). Mobile Financial Services (MFS) comparative summary statement, https://www.bb.org.bd/fnansys/paymentsys/mfsdata.php (accessed: 07.01.2021). Business Finance for the Poor in Bangladesh (2021). A study on Mobile Financial Services for MSEs in Bangladesh: Prospects and Challenges, https://static1.squarespace.com/ static/5a84894df14aa1f379d18290/t/5aa3f91371c10b34ddbdd701/1520695608351/ BFP-B+policy+study+mobile+financial+services.pdf (accessed: 1.11.2021). Chaia, A., Dalal, A., Goland, T., Gonzalez, M.J., Morduch, J., & Schif, R. (2009). Half the World is Unbanked. New York: Financial Access Initiative. Corr, C. (2006). Financial Exclusion in Ireland: An exploratory study and policy review. Dublin: Combat Poverty Agency. Damodaran, A. (2013). Financial Inclusion: Issues and Challenges. AKGEC International Journal of Technology, 4(2), 54-59. Demirguc-Kunt, A., & Klapper, L. (2012). Measuring financial inclusion: The global fin- dex database. World Bank Policy Research Working Paper, 6025, 1-58. Dev, S.M. (2006). Financial Inclusion: Issues and Challenges. Economic and Political Weekly, 41(41), 4310-4313. http://dx.doi.org/10.2307/4418799. Devlin, J.F. (2005). A Detailed Study of Financial Exclusion in the UK. Journal of Consum- er Policy, 28(1), 75-108. http://dx.doi.org/10.1007/s10603-004-7313-y. Duncombe, R., & Boateng, R. (2009). Mobile Phones and Financial Services in De- veloping Countries: a review of concepts, methods, issues, evidence and fu- ture research directions. Third World Quarterly, 30(7), 1237-1258. http://dx.doi. org/10.1080/01436590903134882. Frame, W.S., & White, L.J. (2004). Empirical Studies of Financial Innovation: Lots of Talk, Little Action? Journal of economic literature, 42(1), 116-144. http://dx.doi. org/10.1257/002205104773558065. Granger, C. (1969). Investigating Causal Relations by Econometric Models and Cross-spec- tral Methods. Econometrica, 37(3), 424-438. http://dx.doi.org/10.2307/1912791. Horne, J.C. (1985). Of financial innovations and excesses. The Journal of Finance, 40(3), 620-631. http://dx.doi.org/10.1111/j.1540-6261.1985.tb04984.x. Iqbal, B.A., & Sami, S. (2017). Role of banks in financial inclusion in India. Contaduría y Administración, 62(2), 644-656. http://dx.doi.org/10.1016/j.cya.2017.01.007. Islam, M.T., Talukder, A.K., Siddiqui, M.N., & Islam, T. (2020). Tackling the COVID-19 pan- demic: The Bangladesh perspective. Journal of Public Health Research, 9(4), 389-397. http://dx.doi.org/10.4081/jphr.2020.1794. Johansen, S. (1988). Statistical Analysis of Cointegration Vectors. Journal of Eco- nomic Dynamics and Control, 12(2-3), 231-254. http://dx.doi.org/10.1016/0165- 1889(88)90041-3. Kabir, M.H., Sadrul Huda, S., & Faruq, O. (2020). Mobile Financial Services in the con- text of Bangladesh. Copernican Journal of Finance & Accounting, 9(3), 83-98. http:// dx.doi.org/10.12775/CJFA.2020.013. Kanobe, F., Alexander, P.M., & Bwalya, K.J. (2017). Policies, regulations and proce- dures and their effects on mobile money systems in Uganda. The Elec-tronic Jour- the imPAct of mobile finAnciAl services… 25 nal of Information Systems in Developing Countries, 83(1), 1-15. http://dx.doi. org/10.1002/j.1681-4835.2017.tb00615.x. Kim, D.-W., Yu, J.-S., & Hasan, M.K. (2018). Financial inclusion and economic growth in OIC countries. Research in International Business and Finance, 43, 1-14. http://dx.doi. org/10.1016/j.ribaf.2017.07.178. Kim, M., Zoo, H., Lee, H., & Kang, J. (2018). Mobile financial services, financial inclu- sion, and development:A systematic review of academic literature. The Electronic Journal of Information Systems in Developing Countries, 84(5), 1-17. http://dx.doi. org/10.1002/isd2.12044. Mbogo, M. (2010). The impact of mobile payments on the success and growth of micro- business: The case of M-Pesa in Kenya. Journal of Language, Technology & Entrepre- neurship in Africa, 2(1), 182-203. http://dx.doi.org/10.1057/9781137574633_6. Ozili, P.K. (2018). Impact of digital finance on financial inclusion and stability. Borsa Is- tanbul Review, 18(4), 329-340. http://dx.doi.org/10.2478/fiqf-2020-0023. Pearce, D., & Ortega, C.R. (2012). Financial Inclusion Strategies Reference Framework: World Bank, https://documents1.worldbank.org/curated/en/801151468152092070 /pdf/787610WP0P144500use0only0900A9RD899.pdf (accessed: 02.11.2020). Porteous, D. (2006). The enabling environment for mobile banking in Africa. London: De- partment for International Development. Sarma, M. (2008). Index of Financial Inclusion. ICRIER Working Paper, 215, 1-26. Sarma, M., & Pais, J. (2011). Financial Inclusion and Development. Journal of Internation- al Development, 23(5), 613-628. http://dx.doi.org/10.1002/jid.1698. Sharma, D. (2016). Nexus between financial inclusion and economic growth. Journal of Financial Economic Policy, 8(1), 13-36. http://dx.doi.org/10.1108/JFEP-01-2015- 0004. Siddik, M.N., Sun, G., Yanjuan, C., & Kabiraj, S. (2014). Financial Inclusion through Mobile Banking: A Case of Bangladesh. Journal of Applied Finance & Banking, 4(6), 109-136. Vodafone (2007). The transformational potential of m-Transactions. The Vodafone Pol- icy Paper Series, 6, 1-52. World Bank (2012). Information and Communications for Development 2012: Maximizing mobile. Washington: World Bank Group Publications. World Bank (2012). The little data book on financial inclusion. Washington: World Bank Group Publications. World Bank (2015). How to Measure Financial Inclusion. Washington: World Bank Group Publications. World Bank (2018). The Little Data Book on Financial Inclusion 2018. Washington: World Bank Group Publications.