The difference between early and late
adopters of major fintech innovations.
Name: Tom van der Veen Student number: S4443527
Supervisor: J.Schmitz Date: 15-08-2020
Abstract
E-Banking is a form of IT application that serve customers when doing transactions using their
personal accounts by computer or mobile phone that is connected to the internet, therefore noncash transactions applied. This research aims to analyze the relationship between the application of two major fintech E-bank innovations and loan provision and credit risk for 56 banks out of 19 countries. The data used is sourced mostly from Eikon and the Worldbank and covers data from 1992-2019. As the dependent variables, loan provision is denoted in Total loans, and credit risk is denoted as the ratio of non-performing loans to total loans and the loans to asset ratio. The independent variables are Internet Banking and Mobile Banking applications which are measured using a dummy and early adopters of both technologies which is also measured with a dummy. This study uses bank and country specific control variables and also controls for crises and differences between developing and developed countries. The data analysis technique used are event studies, fixed effects and random effects panel models. The results of the event study indicate that: (1) internet banking and mobile banking applications causes a positive significant effect in the increase of loan provision, (2) internet banking and mobile banking applications do not have any effect on credit risk ratios. The results of the panel regressions of internet banking show that (1) early adopters of internet banking do not issue more loans compared to late adopters, (2) early adopters of internet banking do not have different credit risk ratios compared to late adopters. The results of the panel regressions of mobile banking applications show that (1) early adopters of mobile banking application do issue more loans compared to late adopters, (2) early adopters of internet banking do not have different credit risk ratios compared to late adopters.
List of Content
Abstract ... 2
Introduction ... 4
History of internet banking and mobile banking ... 8
Mobile banking applications ... 9
Literature review ... 11 Internet Banking ... 11 Mobile banking ... 12 Hypotheses development ... 14 Methodology ... 16 Event study ... 17 Panel data ... 19 Dependent variable ... 19 Independent variable ... 20 Control variables ... 20
Panel data model ... 24
Results ... 25
Event study results ... 25
Panel data results ... 28
Conclusion ... 36
References ... 38
Introduction
In the early 1990s the financial markets worldwide were greatly affected by the internet revolution. The advantages aroused by the internet revolution permanently changed the face of financial services and led to the development of many new financial services. Due to the revolution, services such as banking, insurance provision and business transactions are now accessible online without being in physical contact with these services. The development of these services led to new financial business models like online banking, online brokerage services, mobile payment, and mobile banking. According to Lee and Shin (2018) the impact of internet technology has been especially obvious in the banking industry. They argue that many components of the banking business could benefit from the new technologies provided by the internet. In line with this reasoning Nielsen (2002) argues that From the bank’s point of view, potential benefits of online banking include lower operational costs, shorter turnaround time, real-time managerial information, smoother communication within the organization, more convenient interaction with existing as well as prospective customers, and the provision of value-added services such as access to professional knowledge in financial management. When looked upon the consumer’s point of view internet banking services led to the ability to
update account statements, adding to savings, making payments, applying for loans, investing in mutual funds, applying and paying for mortgages, transferring funds and so on (Jayawardhena and Foley, 2000).
Even though, using the internet to access a bank’s services, has been around for a while it is still widely used and is still developing. A more recent example of the development of internet driven technology, is the growth in the usage of mobile banking applications by using smart mobile devices. According to Fenu and Pau (2015), Banks are increasingly investing on mobility, by enabling the mobile web and mobile app channels for online banking, and by providing new mobile payment services. These kind of mobile applications make it possible to manage one’s finances, and allow banks to offer a wider range of services. Other advantages, seen through a bank’s point of view arise from advanced data analysis. Based on user data and transaction histories banks are able to look at everyday decision making of their consumers and hence be able to use this data and mining techniques to identify risky borrowers (Fenu & Pau, 2015). Another benefit of the data is to make tailor offers of new financial services (Vaidya & Diwakar 2008).
Both online banking and mobile banking applications are examples of “FinTech” driven digital innovations disrupting existing industry structures. The term FinTech is a rather simple and obvious combination of an application domain financial and technology (Alt, Beck and Smits, 2018). And previous work on the evolution of FinTech already suggests that financial technologies have a longer legacy than the term FinTech itself (Lee & Shin, 2018). Existing literature, regarding both FinTech
before and after the implementation of the financial innovation. For instance Ceylan, Emre and Deniz (2008) studied the Turkish banking sector during 1996 to 2000 and found that in the first year of adoption usually there is no positive performance observable after implementing online banking. They argue that it often takes two to three years to find a noticeable increase in performance. In addition, the study of Ali and Mousa (2011) analyzes the Jordan banking sector from the year 2000 to 2009 find that it takes time for banks to adopt new e-banking services like online banking, resulting in a negative relationship between a bank’s performance and online banking for the first three years after adoption. On the contrary Okiro and Ndungu (2013) examined the Kenya’s banking sector and found for both internet banking as mobile banking an increase in performance due to increased efficiency, effectiveness and productivity. In addition Cleveland (2016), who invested the top banks of America found that mobile banking applications have a positive effect on a bank’s performance provided that there is a cost reduction in the so called front-line tellers and brick-and-mortar investments.
Existing literature on online banking is subsequently complemented by research towards a possible change in a bank’s risk levels. However for mobile banking applications there is little to no research on this relationship. To determine credit risk, existing literature uses a wide range of determinants like loans to deposit ratio, credit growth, deposit rate and credit rate to identify credit risk (Shkodra amd Ismajli, 2017). Sullivan (2000) was one of the first to compare the loan mix of banks with and without internet banking. His study shows that Internet banks in the Large Regional Bank category tend to have more of their loan portfolio in real estate and consumer loans and less in construction loans. This consumer orientation is consistent with the high adoption rates for retail Internet banking at these banks. Arnaboldi and Claeyes (2008) used a comparative analysis of four different European countries like the UK, Spain, Italy and Finland and look whether there is a difference between pure online banking and mixed banks. They point out that there’s a difference between purely online and mixed banks from the loan perspective. Namely, they argue that pure online banking banks are not so keen into providing loans compared to mixed banks. Onay and Ozsoz (2012) used a panel data in 18 deposit banks in Turkey, in a period of time from 1990-2008 and found a positive relation with online banking on the deposit and credit branch.
After Sullivan (2000) compared the differences in loan provision between banks, he examined if the adoption of internet banking affected performance and risk levels in banks. He argues that banks offering Internet banking are taking a risk by adopting the technology at an early stage of the
product’s life cycle. However he found no noticeable higher risks in banks that were considered early adopters of online banking. Sathye (2005) found similar findings for major credit unions and banks in
towards risks associated with providing credit and is often referred as credit risk. Current literature of the effect of mobile banking applications on credit risk is limited however there are studies use the broad term “e-banking” and the effect on credit risk. An example is the study performed by the Basel Committee which noted that e-banking activities increased and modified some of the traditional risks associated with banking activities, thereby influencing the overall risk profile of banking’ (Basel, 2001). Another study in 2008 published by the world bank suspected an increase of poor, currently unserved people to use mobile banking applications to apply for loans which could lead to an
increase in a bank’s risk preceptors (Ivatury & Mas, 2008).However the provision of more loans could also benefit banks to measure risk exposure, since current account movements, money transfers and payments could be tracked and analyzed. This information gives banks an early warning on whether a client is able to repay a loan (Arnaboldi & Claeys, 2008).
The aim of this study is to compare potential changes in credit risk and loan provision caused by online banking and the introduction of mobile banking applications for 56 banks across 19 countries by using data from 1992 up to 2019. After this comparison is made this study tries to enrich the existing literature by comparing early adopters against late adopters of both services.
To measure a change in loan provision this paper looks at the total loans provided by a bank which represent the total amount of money loaned to customers before reserves for loan losses but after unearned income. And to measure credit risk this paper uses the non-performing loans to total loans ratio and the loans to total asset ratios respectively.
To compare potential changes in credit risk and loan provision caused by fintech innovation this study will use an event study to see if there are any noticeable abnormal returns after implementing a new fintech innovation.
To see if there is any difference between early adopters and late adopters a random effects model is used with a dummy for early adopters as independent variable and various bank specific and country specific control variables. This study also controls for Crisis and differences between developing and developed countries.
The results of the event study indicate that: (1) internet banking and mobile banking applications causes a positive significant effect in the increase of loan provision, (2) internet banking and mobile banking applications do not have any effect on credit risk ratios. The results of the panel regressions of internet banking show that (1) early adopters of internet banking do not issue more loans
compared to late adopters, (2) early adopters of internet banking do not have different credit risk ratios compared to late adopters. The results of the panel regressions of mobile banking applications show that (1) early adopters of mobile banking application do issue more loans compared to late adopters, (2) early adopters of internet banking do not have different credit risk ratios compared to
A study such as this is important for many reasons. Firstly, it fills the research gap by comparing changes in loan provision and credit risk after the rise of mobile banking applications for 56 banks across different countries since this is never done before.
Secondly this study builds on the foundations of earlier research on the adoption of internet banking with the aim to extend the current field by filling a research gap by comparing changes in loan provision and credit risk after the rise of internet banking for 56 banks across different countries. Thirdly, it helps financial institutions in their consideration of applying new financial technologies. Namely, this study will provide insights on possible changes in a bank’s risk perceptions when there is a change in the loan provision induced by adapting new financial innovations.
History of internet banking and mobile banking
Internet banking
Since the existence of banks, banks have made use of the technology available ever since. Banks used and use these technologies primarily to be able to manage growing business volumes. As a side effect the realization of cost reduction due to automation and increasing efficiency established. These side effects definitely existed as an objective but tended to be aimed at the medium to long term perspective (Lamberti & Büger, 2009). Therefore, the relation between banking industry and technology in the 20th and 21th century principally is not a new development, however still a very interesting relation to examine.
The first basic form of a non-branch bank is the the ATM (Automated teller machine). An ATM allows it customers to access their balances, withdraw money and make payments by only using a credit card and a pin to access it. This type of banking is a small machine and could often be found in cities, public buildings and in bank (Haliji, 2014)
However, the first appearance of online banking was in1980 when Home banking was introduced by Citibank, Chase Manhattan, Chemical and Manufactures Hanover. These four banks provided a home banking service were the consumer had to use a terminal, television and keyboard to access the banking system through a phone line. This innovation allowed costumers to take a look at their balances, bank transfers and bill payments (Haliji, 2014). Nowadays online banking looks much different and has become an important technology for costumers everyday life. Over the years many types of online banking were introduced and evolved. The types of online banking which are used mostly throughout history and nowadays are ATM, PC(internet) banking and phone banking (Haliji, 2014).
When Internet banking was first introduced its main use was to provide information of a bank’s products and services. Banks used the Internet to create a website where all kind of products and services were listed for their customers. Due to the technological advancement banks use the Internet for many other reasons like, secured transactions, mortgage applications and so on (Kagan, Acharya, Lingam & Kodepaka, 2005). Internet banking in the early stages helped banks present a potentially low cost alternative to brick and mortar branch banking. According to Burnham (1996) in the early days of Internet banking the majority of banks with Web sites spent less than US$25,000 to create a website, and less than US$25,000 a year maintaining it. Even though the costs of creating and maintaining websites rose it’s still a profitable and cost reducing instrument for banks (Tan & Teo, 2000). Due to the easy access of internet banking by its customers a bank is able to provided
exposure, since current account movements, money transfers and payments could be tracked and analyzed. This information gives banks an early warning on whether a client is able to repay a loan (Arnaboldi & Claeys, 2008).
However, not only banks profit of the introduction of online banking. Especially customers
practicality benefited of internet banking. Major advantages of internet banking are, bank operations are less time consuming, availability of information and a better understanding of a customer’s financial position. For instance Alsajjan & Dennis pointed out that in Finland customers had to travel long distances, spend a lot of time and money to travel to the bank offices. Also as Haliji (2014) points out costumer were able to manage their own accounts due to online banking which gave them more insight in their financial positions. In the end bank customers could now perform common banking transactions such as paying bills, transferring funds printing statements and checking account balances online using only a computer (Acharya and Kagan, 2004).
Mobile banking applications
The innovation of internet banking was quickly followed up by mobile payments. Mobile payments allowed the user to make a money payment for a product or service through a portable electronic device. An example are SMS payments which allowed the costumer to make payments through an SMS text message. A new dimension was added to mobile payments when apple launched their first Iphone. Due to this new mobile device banks were able to launch mobile banking applications which allowed costumers to perform online banking tasks while away from your home computer. These tasks consist out of monitoring account balances, bill payments and transferring funds between accounts. Compared to the traditional banking and even computer based internet banking the introduction of mobile bank offered even more benefits like true freedom from time and place, and efficiency for banking transactions (Laukkanen, 2017).
As stated mobile banking was first introduced in the late 90s-early 2000s when the Internet as a whole began to gain popularity. A few select large banks like Wells Fargo started using the internet to offer rather simple services on their websites. Such services were often limited to checking account balances and finding the nearest ATM. According to Cleveland, (2016) banks like Wells Fargo did not offer interactive services yet. During this period it was unclear in what way the internet would evolve, however many people thought that it may be limited to personal computers. It’s
understandable that the potential of the internet was underestimated and concepts a smart phones were far from thinkable then. Even though smartphones are a part of our normal day lives the mobile banking industry had a rough start according to Cleveland, (2016). Bank websites were accessed
refresh speed and a limited quantity of features contributed to why customers did not readily adopt banking channels beyond physical branches and ATMs. Profitability and investor performance were key aspects to evaluate by both commercial banks and investment banks when investing in mobile banking technology. It was unclear in the early days how current or concept mobile services would interact with the existing offerings, and the employee base, as well as support the bank brands that had been built on personal customer service. The evolution of several technologies thrust mobile banking into mainstream use. The rapid changes in telephone transmission and expanded bandwidth, have allowed platforms to expand and in some cases re-launch mobile banking. Due to other improvements in technologies like data transmission and app development, mobile banking gained popularity around 2009. When banks started with launching their mobile banking applications, and has gained popularity each year in banking institutions offering mobile services. According to the Federal reserve (2015) In 2014, 78% of American financial institutions offered a mobile banking application. As said, technologies are considered a big factor in the gain of popularity for mobile banking applications. However the introductions of the iPhone and other smart phones with similar platforms like android or windows have been a steady factor in the growth of mobile banking. Smart phones have digitized many everyday tasks through the efficiency it brings to everyday tasks allows to regain valuable time spent in applying for loans, transactions and paying bills (Cleveland, 2016).
Mobile banking applications allow banks to find more information about their clients which could be used to develop and design bank services and products specifically aimed at the costumer (Sajić et al, 2017). Another benefit of mobile banking applications is to attract new customers since there are 3.5billion smartphone users worldwide in 2020 (O’Dea, 2020). On the downside research on mobile banking found that common risks associated with mobile banking are fraud and money laundering (Mbawa, 2018).
Literature review
Internet Banking
Existing literature in regard to the adoption of a new fintech innovation is often focused on the impact on a bank’s performance. Studies like Egland et al (1998); Sullivan (2000) found no evidence of differences in the performance of the group of banks offering Internet banking activities compared to those that do not offer such services in terms of profitability, efficiency or credit quality.
On the contrast there are studies who did found evidence for an increase in performance after implementing Internet banking. For instance, Furst et al. (2000a, 2000b, 2002a and 2002b) found that banks in all size categories offering Internet banking were generally more profitable and tended to rely less heavily on traditional banking activities in comparison to non-Internet banks.
In line with Furst et al. (2000a, 200b, 2002a and 200b), DeYoung et al. (2006) observed the change in financial performance of Internet community banks in U.S. during 1999-2001. The results found that Internet adoption improved community banks’ profitability, particularly through increased revenues from deposit service charges.
According to the literature credit risk and loan provision are closely related. Credit risk is specified as the risk arising from a borrower’s failure to meet the terms of a credit contract with the bank or otherwise to perform as agreed. Perumal & Shanmugam (1970) already point out the relation between loans and credit risk before the internet banking industry was born. They argue that Internet banking provides the opportunity for banks to expand their businesses, which leads to an increase in the loan provision. However, it is challenging for institutions to verify their customers as well behaving customers, which is an important element in making sound credit decisions. Verifying customers and their demand for collateral while being as riskless as possible is even more challenging with out-of-area or even overseas borrowers. Hence, unless properly managed, Internet banking and mobile banking applications could lead to a concentration in risky credit provisions.
Previous literature found a connection between loan provision and the overall risk levels of a bank. For instance, Foos, Norden & Weber (2010) investigated how the growth of loan affects the riskiness of banks from 14 western countries in the period 1997-2005. They test 3 hypothesis to show
evidence between loan growth on loan losses, bank provability and bank solvency. They found that loan growth has a positive significant effect on subsequent loan losses with a maximum in the third year. They argue that providing loans at rates which do not compensate for default risk may foster loan growth, however it decreases the bank’s profitability because it decreases the interest income of the average outstanding loan. They also find that loan growth causes capital ratios to fall which decreases the bank’s solvency. The results of this study holds an implication, that banking supervisors
In line with Foos, Norden & Weber (2010), Arnaboldi & Claeys (2008) argues that the information of loan provision could be used to help banks to reduce credit risk exposure, since current account movements, money transfers and payments are currently tracked. This information may be an early warning on clients’ repayment capacity.
Sathye (2005) investigated the impact of transactional internet banking on the risk profile and performance of major credit unions in Australia. They used a censored normal/OLS analysis and used the ratio of provisions for doubtful debts to gross receivables to measure the risk profile as
dependent variable. Sathye (2005) found that there is no significant impact on the risk profile of credit unions. These results could be different compared to banks as credit unions operate under various constraints like they tend to have a smaller size and area to operate in and due to the geographical area of Australia the volume of transactions are limited. The results are still rather interesting since they share similarities with other studies like Sullivan who also found that measure of risk do not point to higher risk in Internet banks when compared to non-Internet banks (Sullivan, 200). However Sathye (2005) does point out that compared to banks credit unions tend to be more prepared to take risks.
Pooja and Balwinder (2009) studied the Indian banking industry and compared internet to non-internet banks over the period 1998-2016. They looked if there is a difference in performance and risk profiles of Internet banks and non-Internet banks. The results showed that there is no significant evidence for performance differences between Internet and non-Internet banks. However they did find that internet banking as a whole had a negative and significant effect on risk. Which indicates that Internet banking has increased the risk profile of all types of banks.
Another study who points out that the adoption of internet banking could actually increase a bank’s risk ratios is the study performed by Pennathur (2001). Pennathur (2001) argues that traditional banking risks such as credit, interest or liquidity risk can worsen if a bank has a significant internet presence.
Mobile banking
Research on the implementation of mobile banking applications compared to internet banking is limited and often focused on customers' adoption, while banks' perspective is neglected. Recently, in their literature review of mobile banking, Shaikh and Karjaluoto (2015) report 55 studies (between 2005 and 2014) associated with different kinds of motivations that influenced mobile banking potential adopters.
When the bank perspective is the main topic of a research it’s often focussed on the performance again. For instance Maina & Mungai (2019) found that mobile banking had a positive effect on the
financial performance of commercial banks in Kenya for the period 2013-2018. They also found a reduction in a bank’s cost because banks could warn their customers with automatic text about loan repayments which saved loan offices time and phone bills. They also argue that due to the
notifications the default rate dropped. In line with Maina & mungai (2019) other studies also found a positive significant increase in performance (Mbawa, 2018; Lasmini et al, 2020; Harelimana, 2017) for respectivally Zimbabwe, Rwanda and Indonesia.
Other existing literature on mobile banking found that mobile-banking applications are aimed at attracting new customers. For instance Simoni (2020) did research on the distribution channels of the banking industry in Albania. He argues that mobile banking has become one of the most important channels for the distribution of banking services like loans. As Simoni (2020) argues that Permanent access, the ability at any time check your balances and make payment, have determined the
popularity of providing loans through this distribution channel of banking services. Crowe et al (2015) investigated the impact of mobile banking for five federal districts in the USA they found that mobile banking is important to customer retention and acquisition goals. They also argue that early adopting of this new service could make an impact, and, at least short-term, advantage may accrue for those early adopters. As future recommendation they recommend to look at difference between large and small financial institutions since their sample mostly consisted out of smaller institutions. Gu et al. (2009) studied consumer trust in mobile banking applications and used one bank in their sample. They argued that attracting potential users and retaining existing users is crucial in the long term of mobile banking applications (Gu et al., 2009). Hence they argue that customer’s profit was the top priority of the bank. Also Fenu & Pau (2015) pointed out the importance of mobile banking for banking institutions at the short term it helps banks in decision making: banks can employ data mining techniques to identify risky borrowers. Secondly it allows banks to tailor offers and new services to customers as a way to keep them or attract new customers (Fenu and Pau, 2015) As for Credit risk Cheng & Qu, (2020). Did research on Fintech innovations for the Chinese banking industry over the period 2008-2017 found that Fintech and found that bank Fintech significantly reduces credit risk in Chinese commercial banks, further results point out a negative effects of bank FinTech on credit risk are relatively weak among large banks, state-owned banks, and listed banks.
Hypotheses development
Consistent with the provided theoretical background
As described in the previous section this study will focus on two different events. Henceforth, this paper will test multiple hypotheses.
H1: The implementation of online banking and mobile banking applications has a positive effect on loan provision.
This hypothesis is supported by findings by for instance Sullivan (2000), who shows that Internet banks in the Large Regional Bank category tend to have more of their loan portfolio in real estate and consumer loans and less in construction loans. This consumer orientation is consistent with the high adoption rates for retail Internet banking at these banks. Also Arnaboldi and Claeyes (2008) used a comparative analysis to look whether there is a difference between pure online banking and mixed banks for four different European countries like the UK, Spain, Italy and Finland .They point out that there’s a difference between purely online and mixed banks from the loan perspective. Namely, they argue that pure online banking banks are not so keen into providing loans compared to mixed banks. The sample of this research mostly consists out of consumer banks(mixed) hence the same results are expected. In line with both studies, Onay and Ozsoz (2012) used a panel data in 18 deposit banks in Turkey, in a period of time from 1990-2008 and found a positive relation with online banking on the deposit and credit branch.
As for an increase in loan provision due to mobile banking applications studies like those of Simoni (2020); Gu et al (2009); Fenu and Pau (2015) and Crowe et al(2015) argued that mobile banking application are aimed at attracting and retaining existing customers. They found evidence for this link in the United states, Albanian and south Korean banking industry.
H2: The implementation of online banking has no effect on a bank’s credit risk.
This hypothesis is in line with the findings of Sullivan (2000) and Sathye (2005) who respectively studied banks in the tenth federal reserve district of the United States and Credit unions in Australia. Both found no effect on credit risk due to internet adoption.
H3: The implementation of mobile banking application has a positive effect on a bank’s credit risk. In 2001 the Basel committee researched the possible downsides of mobile banking. They argued that mobile banking could cause an increase in so called “bad loans” due to the increase in new
customers which couldn’t get a loan before. However more recent research towards this like Fenu and Pau (2015) show that banks could reduce credit risk because they gain more information on their costumers due to mobile banking applications. Also Cheng & Qu, (2020) found a significant reduction in credit risk for banks using Fintech innovations like mobile banking applications.
H4: There is a difference in loan provision between late and early adopters of internet banking and mobile banking applications.
Sullivan (2000) argues that in the early stage of introduction of internet banking there is
considerable uncertainty about the demand for a new product. He also points out that according to financial analysts due to this uncertainty large banks dominate the internet banking market in the early stages of internet banking. The reason for this is because these banks have more capital buffers and are hence able to absorb the cost of introducing and developing the product. Smaller banks are henceforth more likely to be late adopters of the product and will jump in when there is a certain demand for the product. Since early adopters are more likely to dominate the market it is expected that early adopters are more likely to show an increase in loan distribution.
This relationship is also shown by Nath et al (2001) who collected data of 75 banks and customer respondents. The Respondent felt that banks not providing internet banking would potentially lose customers to competitors who offer such services. This idea is shared by the bank respondents since a large part of the sample who did not introduced internet banking yet are planning to do so in the near future. They also expected, based on the bank respondents feedback, that internet banking offers a wide range of new options that could be exploit by banks causing an increase in the number of customer accounts.
H5: There is no difference in credit risk ratio’s between late and early adopters of internet banking and mobile banking
According to Sullivan (2000), banks offering Internet banking are taking a risk by adopting the technology at an early stage of the product’s life cycle. They may also be willing to generally accept more risk compared to late adopters. However, he find that banks have been neither helped nor harmed by their early commitment to the Internet as a delivery channel.
So far there has not been done any research on early adopters of mobile banking which is focussed from a bank’s perspective. So the hypothesis regarding credit risk shall be the same for mobile banking applications as it is for internet banking.
Methodology
This paper collects the data from 1992 to 2019 of 56 banks who implemented both internet banking and mobile banking applications out of twenty countries. From this data this paper fist examines whether there is an effect between the adoption of the two fintech innovations and a bank’s loan provision and credit risk. To give an insight in this effect firstly this paper will use an event study to see if there are any abnormal returns in loan provision and the credit ratio of the bank. An event study is used because all banks in the sample implemented the fintech innovation so there is no “control group” like non-internet banks, which is often used in other studies. After the event study the paper will use both a fixed and random effects model to answer the research question if there is a difference between early and late adopters. According to Furst et al (2002) in the early days of internet banking the costumer usage remained low which makes it unlikely that Internet banking is having a sizeable impact on most banks. However as they argue, an exception to this generalization are most of the considered large banks of each country. As they argue that all of the largest banks in these countries offered Internet banking, but only about 7 percent of the smallest banks offered it (Furst et al, 2002). Nowadays most of the banks use internet banking, that is why this paper uses both large and small banks to see if there is any difference.
The same approach will be used to see if there is any difference between early and late adopters of mobile banking applications.
The data will be collected from different sources. The main data source for this study is Refinitv Eikon which will be used to collect the data on banks. For the country specific control variables this study uses data from the world bank and to find the dates on which banks adopted either internet banking and mobile banking applications annual reports and news articles will be used.
Event study
Existing literature often makes use of event studies to see if there is a reaction when a certain event occurs (Temming, 2014; Delianedis and Geske, 2003; Norden and Weber, 2004). For instance, Temming (2014) studies the effect of M&A announcements on the stock price or Delianedis and Geske (2003) who studied the effect of risk neutral probabilities or default announcements on credit risk. Even though this paper will not examine the effect of an event on the stock price, the same method could be implied to examine the effect of major Fintech announcements on an abnormal increase in the provision of loans and credit risk. And henceforth it could be used to test the first, second and third hypothesis. Using event studies this study is able to compare several sub-samples like different time intervals for event windows to discover the effect of announcements regarding the implementation of online banking and mobile banking applications. Another usage of using an event study is to compare possible abnormal increases in loan provision for both announcements. Since there is no existing event study on the effect of an abnormal increase in loans due to a Fintech announcement, the “optimal” event window is determined in the process. When performing an event study you have to follow a few steps.
(i) Defining the event
Defining the event is basically nothing more than to find a particular date an event occurs. In this study the event differs per bank and has to be looked up individually. For instance Wells Fagro launched Internet banking in May of 1995 whereas the Royal bank of Canada launched their internet banking in October of 1997. These event dates do also differ for adapting mobile banking
applications.
(ii) Estimation procedure
As shown in Figure 1 above the event study consists out of an estimation window and event window. This study will use for both internet banking and mobile banking applications an estimation window of 3 year prior to the event (0,-3). The event study will exist out of different event windows which are (0,+1) (0,+3) (+1,+3) (+3,+5). All steps are in years. For estimation windows there is no rule of thumb that could be followed. Most estimation windows are around what prior studies did. Since most studies regarding internet banking use windows prior, around and after the 3 year mark this study will do so to. These window will also be used for mobile banking. As for the estimation window Armitage (1995) suggest to use a period of 24 to 60 months when performing an event study with monthly data. Different estimation windows were created however the most appropriate one was 36 months prior to the events. The However, as most literature on event studies do, this study will compare the time before the event with the time afterwards.
(iii) The creation of the index
The calculated abnormal returns should be regressed on a given index. This study uses the average net loan, non-performing loans to total loans and loan to asset ratio for each individual bank to create an index. The reason we didn’t use the average of all banks for each individual variable is because of the small sample size and difference in size of each individual bank .
(iv) Defining predicted return, abnormal return and cumulative abnormal return
𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡(𝑎𝑖+ 𝐵𝑖𝑅𝑖𝑡) (5)
Where:
𝐴𝑅𝑖𝑡: abnormal return
𝑅𝑖𝑡: actual return of the different event windows (𝑎𝑖+ 𝐵𝑖𝑅𝑖𝑡): Predicted normal return by the index
𝐶𝐴𝑅𝑖(𝑡1− 𝑡2) = ∑ 𝐴𝑅𝑖𝑡 𝑡=2
𝑡=𝑡1
(6)
Where:
Panel data
In order to test the hypothesis 4 and 5 we’ve to examine the panel effect. Since the sample of this study consist out of banks across various countries the study wants to examine the effect between these countries, hence panel regression will be used. A simple OLS or cross sectional model could not be used since the event windows are larger than one year and the sample exists out of banks out of multiple countries. Hence when using this kind of data, most studies use a fixed-effect model or a random-effect model. The random effects model has a potential disadvantage which is the chance that there is a correlation between the explanatory variables and the error terms which results in biased parameters. The disadvantage of the fixed effects model is when variables do not differ over time, they cannot be measured. A fixed effects model is problematic because this study wants to examine the effect between early and late adopters of the two fintech innovations. This variable, and others, do not differ over time and could not be measured with a fixed effects model. Hence, to answer the research question this study will make use of a random effects model. However to get a better understanding of which bank related variables could potentially influence total loans or credit risk this paper will also examine a fixed effects model with an implementation dummy variable.
Dependent variable
The dependent variables used in this paper are the loan provision and a bank’s credit risk. Loan provision is denoted as:
𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑎𝑛𝑠𝑖𝑡 (1)
𝑖 = 𝑏𝑎𝑛𝑘 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟, 𝑡 = 𝑡𝑖𝑚𝑒 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟
The other dependent variable is credit risk. The reason this paper observes credit risk is explained by Al-Smadi (2011). He argues that credit risk is the most important risk factor for measuring a bank’s risk exposure in e-banking. Hence, whenever an increase in credit risk is observed this is normally associated with a decrease in a bank’s profitability.
There are multiple ways to measure credit risk, one measure of credit risk is a bank’s loan-to-asset ratio. This ratio implies that the more loans a bank makes, the more exposure it has to bad loans. This method is used by for instance Sulivan, 2000; Haliji, 2014;
Another way to measure credit risk is using the ratio of nonperforming loans to total loans. When this ratio is used, a negative sign for this variable is expected (Al-Smadi, 2011; Ahmad, 2020). This study will use both types of credit risk measurements to get a better understanding in which way a bank’s credit risk is affected. The formulations of every credit risk are listed below.
𝑁𝑜𝑛 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝐿𝑜𝑎𝑛𝑠 𝑡𝑜 𝑇𝑜𝑎𝑙 𝐿𝑜𝑎𝑛𝑠𝑖𝑡 =
𝑁𝑜𝑛 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝐿𝑜𝑎𝑛𝑠𝑖𝑡 𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑎𝑛𝑠𝑖𝑡
(3)
Independent variable
Since the data set consists only out of banks who adopted internet banking and mobile banking applications, the independent variables used in this paper is whether a country is a late or early adopter of the specific service. Early adopters of internet banking are the banks that implemented internet banking before 1999. Banks are considered late adopters when they introduced internet banking after 1999. Early adopters of mobile banking applications are the banks that implemented internet banking before 2011. Banks are considered late adopters when they introduced internet banking after 2011. This variable will be captured by a dummy whereas early adopter=1 and late adopter=0
Control variables
As control variables this study will use bank specific and country specific control variables. For each dependent variable different control variables will be used. The control variables that have bank-specific factors and are related to total loans are Loans to Asset ratio which is denoted in formula 2, Non-performing loans to total loans which is denoted in formula 3. As discussed in the literature review the levels of credit risk could influence the total loans issued by banks. Hence it’s of importance to control for this variable. And since this study uses two measurements of credit risk, both will be used as control variable for total loans issued. Both risk ratios should be positively related to total loans because the more credit risk is expected when a banks has more loans. Another control variables are Total Deposits which should be positively related to Total loans. When a bank has more deposits it is able to issue more loans.
The control variables that have bank-specific factors and are related to credit risk are based on Ahmad (2020). He studied the determinants of credit risk on a multi-country bank sample. According to Ahmad (2020) the following variables might influence a bank’s credit risk ratio :
management efficiency 𝑖𝑡 =𝐸𝑎𝑟𝑛𝑖𝑛𝑔 𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡
(2)
loan to − deposit ratio𝑖𝑡=
𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑎𝑛𝑠𝑖𝑡 𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑖𝑡
𝐹𝑢𝑛𝑑𝑖𝑛𝑔 𝑐𝑜𝑠𝑡𝑠𝑖𝑡 =
𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑒𝑥𝑝𝑒𝑛𝑠𝑒𝑖𝑡− 𝑛𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑒𝑥𝑝𝑒𝑛𝑠𝑒𝑖𝑡 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡
(2)
Other control variables which are used in both models are bank size, employment growth, and crisis dummies. Bank size can be reflected in the number of assets owned by the company. The large amount of assets reflects that the bank’s activities such as loan provision could also increase (Sathye, 2005). Moreover, it could be expected that bank size has a positive effect on bank’s loan provision. Size could also be related to credit risk. A relative large bank tends to have a bigger buffer against loan defaults compared to smaller bank so it is expected that a large bank has less credit risk. This line of reasoning is studied by Stever (2007), who found that Small banks tend to be more risky in their loan portfolio because they cannot diversify away idiosyncratic volatility as well as large bank. This inability of small banks to diversify comes about a number of different ways, as Stever argues for example; when a bank has less total loans held and less diversity in borrower type, since it is not able to attract large borrowers, combined with geographic restrictions small banks tend to be more risky compare to larger banks. This study will denote company size as the logarithm of the total assets. This study also controls for the Dotcom bubble (2000-2002) Since most banks in our sample adopted or just adopted internet banking during this period it is important to control for. Also when looking at the total loans provided during this period there is a decrease in total loans for banks in Germany, United States and Singapore Figure 3 in the appendix APPENDIX LINK. This decrease is most likely caused by the dotcom bubble and hence controlled for. The Dotcom bubble will be defined as a dummy whereas the years 2000-2002=1 and all other years=0
Another Crisis this study controls for is the financial crisis of 2007-2008. When looked upon the total loans during this period Australia, Canada, United Kingdom and the United states show a decrease in loan distribution Figure 3 in the appendix. This decrease is most likely caused by the financial crisis and henceforth controlled for. Credit crisis is defined as a dummy and equals 1 for the years 2007-2008 and 0 otherwise.
To control for country specific which are related to both credit risk and loan provision this study uses two macroeconomic variables which are Inflation and economic growth. To measure economic growth this study will use the growth rate of a country’s GDP per capita. According Brissimis, and Delis (2008) credit risk could be affected during economic slowdowns because lending could decrease and vice versa. Another reason this study uses these variables is because they are widely
used when controlling on the effect on credit risk (Al-Smadi, 2011; Foos, Norden & Weber, 2010; Jesus & Gabriel, 2006).
The assumption behind the existing literature on innovations is that it is universally applicable across the world. According to Vaghjiani (2010) this assumption is false. He studied the Australian and Indian banking sector and looked whether IT innovations differ between the countries. He choose internet banking and mobile banking as innovations to study. According to Vaghjiani (2010) factors which influence banks to introduce internet banking in developed countries include cost savings and consumers’ convenience, while developing countries aim to increase the number of consumers (AL-Hajri & Tainall 2011). Therefore it is expected that the motivation between the introduction of internet banking and mobile banking applications is dissimilar between developed and developing countries. With different motivations to introduce a fintech innovation it is likely that the results differ between developing and developed countries. This study will use a dummy variable to capture these effects. Whether a country is considered developed or developing relies on the country classification of the UN. A graphical overview of the total loans issued by both developed and developing countries is showed in the graph below.
The final control country specific control variable is Internet access. When internet access increases it’s likely more people will use internet banking and apply for loans. Figure 2 gives an overview of the percentage of people with access to the internet per country.
Figure 2
Table 1 shows the summary statistics for all variables used. Variables like GDP and Internet access show a large difference in their dataset. This could be explained due to the fact that both grew rapidly the last thirty years. In the Apendix of this paper Table 12 gives an overview of all the Variable definitions and their sources. Table 13 gives an overview of the sample and implementation dates of each innovation. and Table 14 shows the correlation estimations of all these variables for all
dependent variable regressions. Table 1. Summary statistics
(1)
(2)
(3)
(4)
(5)
VARIABLES
N
mean
sd
min
max
Totalloans1
56
1.737e+08
2.169e+08
48,775
1.802e+09
TotalDeposits1
56
1.653e+08
2.358e+08
76,915
2.025e+09
Employeegrowth1
56
4.428
14.31
-50.98
98.58
InternetAccess1
56
41.38
33.90
0.000111
97.32
Inflation1
56
3.674
5.009
-1.736
66.01
NPLtoTotalLoans1
56
3.867
6.693
0.186
62.38
LoanstoAsset1
56
61.48
10.72
26.41
84.85
Earningassetstotot1
56
86.43
4.996
66.54
96.50
Funding_costs
56
0.0643
0.0534
0.0204
0.811
Loan_to_deposit_Ratio
56
1.192
0.830
0.288
25.26
Size
56
20.76
2.076
16.78
27.53
Panel data model
Combining the dependent variables, independent and control variables, this study will use the following fixed effects models :
𝐶𝑟𝑒𝑑𝑖𝑡 𝑟𝑖𝑠𝑘𝑖𝑡= 𝛽0+ 𝛽1𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖𝑡+ 𝛽2𝐿𝐷𝑖𝑡+ 𝛽3𝑀𝐸𝑖𝑡+ 𝛽5𝐹𝐶𝑂𝑆𝑇𝑖𝑡+ 𝛽6𝐸𝑚𝑝𝑖𝑡+ 𝛽7𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽8𝐷𝑢𝑚𝑚𝑦𝐶𝑟𝑖𝑠𝑖𝑠𝑖𝑡 + 𝛽9𝐷𝑢𝑚𝑚𝑦𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑖𝑡+ 𝛽10𝐼𝐴𝑗𝑡+ 𝛽11𝐺𝐷𝑃𝑗𝑡+ 𝛽12𝐼𝑁𝐹𝑗𝑡+ 𝜇𝑖𝑡 (7) 𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠𝑖𝑡= 𝛽0+ 𝛽1𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖𝑡+ 𝛽2𝐿𝑇𝐴𝑖𝑡+ 𝛽3𝑁𝑃𝐿𝑖𝑡+ 𝛽4𝑇𝑜𝑡𝑎𝑙𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑖𝑡+ 𝛽5𝑀𝐸𝑖𝑡+ 𝛽6𝐸𝑚𝑝𝑖𝑡 + 𝛽7𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽8𝐷𝑢𝑚𝑚𝑦𝐷𝐵𝑖𝑡+ 𝛽9𝐷𝑢𝑚𝑚𝑦𝐷𝐶𝑖𝑡+ 𝛽10𝐼𝐴𝑗𝑡+ 𝛽11𝐺𝐷𝑃𝑗𝑡+ 𝛽12𝐼𝑁𝐹𝑗𝑡+ 𝜇𝑗𝑡+ 𝜖𝑖𝑡 (8)
Random effects models:
𝐶𝑟𝑒𝑑𝑖𝑡 𝑟𝑖𝑠𝑘𝑖𝑡= 𝛽0+ 𝛽1(𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖𝑡− ∅𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖) + 𝛽2(𝐿𝐷𝑖𝑡− ∅𝐿𝐷𝑖) + 𝛽3(𝑀𝐸𝑖𝑡− ∅𝑀𝐸𝑖) + + 𝛽5(𝐹𝐶𝑂𝑆𝑇𝑖𝑡− ∅𝐹𝐶𝑂𝑆𝑇𝑖) + 𝛽6(𝐸𝑚𝑝𝑖𝑡− ∅𝐸𝑚𝑝𝑖) + 𝛽7(𝑆𝑖𝑧𝑒𝑖𝑡− ∅𝑆𝑖𝑧𝑒𝑖) + 𝛽8(𝐷𝑢𝑚𝑚𝑦𝐶𝑟𝑖𝑠𝑖𝑠𝑖𝑡 − ∅𝐷𝑢𝑚𝑚𝑦𝐶𝑟𝑖𝑠𝑖𝑠𝑖) + 𝛽9(𝐷𝑢𝑚𝑚𝑦𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑗𝑡− ∅𝐷𝑢𝑚𝑚𝑦𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑗) + 𝛽10(𝐼𝐴𝑗𝑡− ∅𝐼𝐴𝑗) + 𝛽11(𝐺𝐷𝑃𝑗𝑡− ∅𝐺𝐷𝑃𝑗) + 𝛽12(𝐼𝑁𝐹𝑗𝑡− ∅𝐼𝑁𝐹𝑗) + 𝜇𝑖𝑗+ 𝜖𝑖𝑡 (9) 𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠𝑖𝑡= 𝛽0+ 𝛽1(𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖𝑡− ∅𝐷𝑢𝑚𝑚𝑦𝐸𝑎𝑟𝑙𝑦𝑖) + 𝛽2(𝐿𝑇𝐴𝑖𝑡− ∅𝐿𝑇𝐴𝑖) + 𝛽3(𝑁𝑃𝐿𝑖𝑡− ∅𝑁𝑃𝐿𝑖) + 𝛽4(𝑇𝑜𝑡𝑎𝑙𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑖𝑡− ∅𝑇𝑜𝑡𝑎𝑙𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑖) + 𝛽5(𝑀𝐸𝑖𝑡− ∅𝑀𝐸𝑖) + 𝛽6(𝐸𝑚𝑝𝑖𝑡− ∅𝐸𝑚𝑝𝑖) + 𝛽7(𝑆𝑖𝑧𝑒𝑖𝑡− ∅𝑆𝑖𝑧𝑒𝑖) + 𝛽8(𝐷𝑢𝑚𝑚𝑦𝐶𝑟𝑖𝑠𝑖𝑠𝑖𝑡− ∅𝐷𝑢𝑚𝑚𝑦𝐶𝑟𝑖𝑠𝑖𝑠𝑖) + 𝛽9(𝐷𝑢𝑚𝑚𝑦𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑗𝑡 − ∅𝐷𝑢𝑚𝑚𝑦𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑗) + 𝛽10(𝐼𝐴𝑗𝑡− ∅𝐼𝐴𝑗) + 𝛽11(𝐺𝐷𝑃𝑗𝑡− ∅𝐺𝐷𝑃𝑗) + 𝛽12(𝐼𝑁𝐹𝑗𝑡− ∅𝐼𝑁𝐹𝑗) + 𝜇𝑖𝑗 + 𝜖𝑖𝑡 (10) Where:
𝐿𝐷𝑖𝑡: Loan to deposit ratio ME𝑖𝑡: Manager efficiency 𝐹𝐶𝑂𝑆𝑇𝑖𝑡: Funding costs 𝐸𝑚𝑝𝑖𝑡: Employment growth 𝐼𝐴𝑗𝑡:Internet access
𝐿𝑇𝐴𝑖𝑡: Loan to assets ratio
𝑁𝑃𝐿𝑖𝑡: Non-performing loans to total loans 𝜇𝑖𝑡 = country specific error
𝜖𝑖𝑡 = bank specific error
To test which model is the most feasible for each dependent variable an Hausman test will be performed. The results of the Hausman test are listedinTable 15 Hausman testsTable 15 of the appendix. However as this study wants to study difference between and within banks both the fixed effects and random effects models will be used.
Results
In this section, the results of the research will be described. First of the results of the event study for each event window will be shown and explained using a cumulative abnormal return approach which fits with the existing literature. The event study will be used to show if there is a relationship
between the implementation of internet banking and mobile banking. After the event study a panel data analyses will be conducted and explained.
Event study results
The event study analyses will be used to test the first two hypothesis of this paper. Hence, the event study examines the effect of both the internet banking and mobile banking application adoption on total loans and credit risk, using multiple event windows for the three dependent variables which are already specified in the methodological section of this paper. The best window used in this study will be determined after analyzing the results. As described in the methodological part of this paper the abnormal returns for each variable will be calculated and used to create the cumulative abnormal returns. The cumulative abnormal returns are used to give a more general image on whether there are abnormal returns in a given time window. Table 2 shows the cumulative abnormal returns for total loans after the internet banking event. The event study consists out of four periods as described in the method section. All generated event windows shown in Table 2show very positive significant results suggesting there is an increase in a bank’s total loan provision for each event windows after the adoption of internet banking. This means that there is a positive significant increase in total loans after a bank launches Internet banking. A possible reason behind the significant increase of total loans is as Sullivan (2000) argues due to internet banking allowing all banks to easily offer innovative products and access new customers.
Table 2. Cumulative abnormal returns Total loans for event “Internet banking”
Date
Mean
t-statistic
CAR (0,1)
2.28e+08
5.27***
CAR (0,3)
7.36e+08
5.88 ***
CAR (1,3)
9.64e+08
5.20***
CAR (3,5)
1.26e+09
4.35***
*** p<0.01, ** p<0.05, * p<0.1
Table 3 shows the cumulative abnormal returns for the first measurement of Credit risk which is the loan to asset ratio. The results show that there is a positive relation between internet banking and credit risk in year one and from three to five years after adopting internet banking. However the results provided in Table 3 show no significant effects on credit risk. These findings are in line with
previous studies conducted by Sathye (2005) and Sullivan (2000) who didn’t found a change in credit risk after implementing online banking.
Table 3. Cumulative abnormal returns Loans to Assets for event “Internet banking”
Date
Mean
t-statistic
CAR (0,1)
1.271296
0.22
CAR (0,3)
-3.379459
-0.34
CAR (1,3)
-2.108164
-0.15
CAR (3,5)
1.107134
0.12
*** p<0.01, ** p<0.05, * p<0.1
Table 4 shows the cumulative abnormal returns for the second measurement of credit risk which is the Non-performing loans to total loans ratio. Again none of the event windows show any significant results. Which means that there is no increase or decrease in credit risk after banks implemented internet banking.
Table 4. Cumulative abnormal returns Non-performing loans to total loans for event “Internet banking”
Date
Mean
t-statistic
CAR (0,1)
-5.947514
-1.01
CAR (0,3)
-1.634484
-0.17
CAR (1,3)
-7.581997
-0.78
CAR (3,5)
.4242397
0.39
*** p<0.01, ** p<0.05, * p<0.1
Table 5shows the cumulative abnormal returns for total loans after the internet banking event. All generated event windows shown in Table 5. show very positive significant results suggesting there is an increase in a bank’s total loan provision for each event windows after the introduction of mobile banking applications.
Table 5Cumulative abnormal returns Total loans for event “mobile banking applications”
Date
Mean
t-statistic
CAR (0,1)
2.66e+08
2.55***
CAR (0,3)
1.25e+09
3.15***
CAR (1,3)
8.27e+08
3.57***
CAR (3,5)
1.32e+09
3.30***
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 6 and Table 7show respectively the cumulative abnormal return for both credit risk ratios. The periods after the event show various positive and negative increases in these ratios. However none of them are significantly different from the mean. Hence it could be concluded that there are no
abnormal returns in credit risk ratios after implementing either of the 2 fintech technologies.
Table 6Cumulative abnormal returns Loans to Assets for event “mobile banking applications”
Date
Mean
t-statistic
CAR (0,1)
5.317343
0.91
CAR (0,3)
-8.91552
-0.53
CAR (1,3)
-14.23286
-1.06
CAR (3,5)
-8.961787
-0.85
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 7 Cumulative abnormal returns Non-performing loans to total loans for event “mobile banking applications”
Date
Mean
t-statistic
CAR (0,1)
-5.120143
-1.57
CAR (0,3)
-.4975791
-0.06
CAR (1,3)
4.622565
0.55
CAR (3,5)
8.352063
1.15
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Panel data results
As observed with the event study there is a link between internet banking and mobile banking applications on loan provision. The event study also gave insight on the link between internet banking and mobile banking applications on credit risk. However during this period many things could have happened which could have influence either the loan provision and credit risk ratios. To test if hypothesis 4 and 5 hold different panel regressions will be used.
In Table 8 a fixed effects model is displayed which uses a dummy for internet banking as independent variable. The fixed effects models could give insights on what affects loan provision between banks. The results of the first fixed effects regression (1) show that there is a positive relation between the adoption of internet banking and loan provision. However this result is not statistically significant. Internet access on the other hand is positively significant at the 1% confidence interval which means that when there is an increase in internet access a bank’s loan provision increases with 947,771 for each percent of internet users. Also GDP per capita shows a positive significant relation at 1%
confidence interval level with loan provision. The dot com bubble also shows a very strong significant relation on the 1% level which is not in line with what is expected. However since the sample also consists out of developing countries, who possibly, suffered less from the dot com bubble might have influenced the results. All other variable show significant relation with the provision of loans and the R-squared is 0.545 which means it predicts loan provision quite well.
Next a fixed effects regression is used for the NPL to total loans ratio (2) with an internet banking dummy as independent variable. A negative relation between internet banking and credit risk is found however there is no significant effect between internet banking and credit risk. Here there is only a positive significant effect at 5% confidence interval between funding cost and credit risk which means that when interest expenses go up or total assets go down, credit risk rises. This regression does a rather poor job in explaining the growth of credit risk since the R-squared is extremely small (0.043). The last fixed effects model uses the loan to asset ratio as dependent variable and again doesn’t find any significant relation between internet banking and credit risk. However here size has a strong negative effect on the credit risk which means that whenever a bank size increases the credit risk goes down. This makes sense because larger banks tend to have more options to diversify their risks which is in line with the findings of Stever (2007) who argues that larger banks are better in diversifying their risks. This last model does a decent job in explain the growth of credit risk since the R-squared is 0.26.
Table 8 Fixed effects models for internet banking
(1)
(2)
(3)
VARIABLES
Fixed Effects
Fixed Effects
Fixed Effects
Internet banking dummy
1.771e+06
-0.653
-0.428
(6.442e+06)
(0.482)
(0.487)
Loan to deposit
-0.386
-0.161
(0.594)
(0.608)
Employee growth
36,979
0.0164
0.00731
(136,826)
(0.00998)
(0.0101)
Earning assets to total assets
-151,406
0.00975
(1.141e+06)
(0.0774)
Funding costs
38.68**
29.11*
(16.84)
(16.68)
Size
-1.471e+07
-0.601
-4.226***
(1.004e+07)
(0.750)
(0.766)
Internet Access
947,771***
0.0205
0.00878
(211,029)
(0.0155)
(0.0157)
GDP(capita)
8,599***
1.00e-05
5.60e-05
(892.5)
(6.67e-05)
(6.78e-05)
Inflation
-139,881
-0.0176
-0.0307
(370,702)
(0.0278)
(0.0285)
DotcomBubble
3.154e+07***
-0.506
0.199
(6.757e+06)
(0.502)
(0.512)
Loans to Asset
-165,629
(671,499)
NPLtoTotalLoans
256,345
(692,574)
Constant
2.209e+08
13.38
80.44***
(2.156e+08)
(16.60)
(17.02)
Observations
455
447
458
R-squared
0.545
0.043
0.257
Number of id
52
52
53
Country FE
YES
YES
YES
Standard errors in parentheses
*significance at the 10 percent level
** significance at the 5 percent level
*** significance at the 1 percent level
Fixed effect regressions for adoption of internet banking for the period 1992-2002
(1) Fixed effects regression for Total loans as dependent variable
(2) Fixed effects regression for NPL to total loans as dependent variable
(3) Fixed effects regression for Loans to assets as dependent variable
Next this study wants to control for variables which remain constant over time to test if there is a difference in loan provision between late and early adopters of internet banking and mobile banking applications. Hence a random effects model is used since the early adopter dummy remains constant over time and will be omitted by a fixed effects regression. Table 9 shows the random effects
analyses of the three dependent variables whereas (1) is loan provision (2) NPL to total loan ratio and (3) the loan to assets ratio. The first (1) regression shows a negative relation between early adopters and loan provision however this relation isn’t significant so there is no real effect going on here. Which means that the fourth hypothesis doesn’t hold and has to be rejected.
The analyses however does find a significant positive relation at the 1% level for total deposits. The random effects model, as the fixed effects model of Table 8, finds a positive relation between internet access and GDP per capita on loan provision. However the relation between internet access is rather weak since it’s only significant at the 1% level.
The random effects models in Table 9, just like the fixed effects model in Table 8 don’t find a relation between early adopters of online banking and credit risk. This is in line with the fifth hypothesis which states that there is no difference in credit risk between early and late adopters of online banking. So the hypothesis holds for now. Credit risk ratio (2) is again effected by the funding costs, which shows a positive relation on the 5% confidence interval level. Almost an identical relation is found between credit risk ratio (3) and funding cost which is also positive significant at the 5% confidence interval. However compared to the fixed effects model there is also a relationship going on between employee growth and Size for credit risk ratio (2). Employee growth has a very small positive significant effect on credit risk at the 10% level which means that whenever there is an increase in employees credit risk slightly increases. Size has a positive negative for both credit risk measurements which is again in line with Stever (2007). The last variable which has an effect on credit risk (3) is the developed country dummy which means that developed country have less credit risk compared to developing countries. A Robustness check can be found in the Apendix in Table 16 for all three random effects regression. However the results remain similar to the results in Table 9.
Table 9. Random effects model for internet banking
