The effect of age on the sense of security in internet
banking. A study on Dutch internet banking users.
Name: Mick de Nijs Supervisor name: Dhr. M.A. Dijkstra
Student ID: 10253092 Specialization: Finance
This paper examines the drivers of the sense of security of internet banking users. The objective is to find the effect of age on sense of security. By conducting a survey, information is gathered by asking questions regarding peoples sense of security during internet banking activities, age, gender, monthly income, internet banking frequency of use, education level and use of bank. This information is used to conduct an ordinary least square regression to (1) analyze the response variable sense of security and (2) analyze the response variable frequency of usage. Also, a logistic regression is conducted to analyze the effect of age on mobile usage because this is assumed to be related to the sense of security of people. Results show that older people to feel significantly less safe while internet banking and that people who use internet banking often, in other words, have a high frequency of usage, feel significantly safer. Individuals with a high income also feel more safe while internet banking, but do not internet bank more often. A regression to analyze the determinants of frequency of use has been done and it can be concluded that sense of security significantly affects the frequency of usage positively.
1. Introduction
In april 2013, cyber-attacks has been launched on the Dutch bank ING. As a result of the attacks, people might feel less secure about the online banking system. On top of that, Hans Hagenaars, manager at ING bank , recently announced that he wants ING to implement the sale of client payment behaviour to companies. This announcement made people aware of the fact that their privacy information can be easily distributed towards third parties. This awareness and the danger of cyber-threats might affect their usage of internet banking.
Expected security is a significant quality attribute in internet banking (Liao and Cheung ,2002). Speece and Rotchanakitumnuai (2003) argue that the adoption and continued usage of internet banking is negatively related to security concerns. Kumbhar, (2011); Akinci et al,(2004) and Matilla et al, (2003) argued that demographic and geographical factors influenced the adoption of internet banking. The age of the customer has negative relations to the adoption and usage of internet banking. It is interesting to investigate whether this negative relation comes from the argument of Speece and Rotchanakitumnuai (2003) that age is negatively related to security concerns. Banks can then target older people in an attempt to increase their sense of security. This raises the question whether the age of an individual has an effect on their sense of security while internet banking. The research questions is:
How does age affect the sense of security amongst Dutch internet banking users?
Chin et al. 2012 argued that people have more security concerns for internet banking on their mobile phone (m-banking) compared to their computer. Therefore, this research also focusses on the question whether age has an effect on the usage of mobile phones for internet banking purposes. A regression on mobile usage is used to proxy the sense of security in an objective manner.
For banks it is important to know what determines peoples frequency of use of internet banking because then they can adjust themselves and increase usage. To find out whether young or old individuals needs to be targeted to increase usage the following sub research question has been stated:
How does age affects the frequency of usage amongst Dutch internet banking users?
These research questions will be analysed by using information gathered from a survey amongst 50 internet banking users. This information is used in three regressions. The first linear regression tests the effects on sense of security. The second linear regression examines the effects on frequency of usage. Results show that sense of security positively affects frequency of usage, therefore, it is for banks important to reduce security concerns, so
that customers will use internet banking services more frequent. At last, a logistic regression has been conducted to analyse effects on mobile usage. This has been done because mobile banking is relatively new and is related to sense of security as shown in this article.
The rest of this thesis is organized as follows : Part 2 and 3 consist of a literature and empirical section where the aim is to explain which factors are of importance towards customers internet banking behaviour. Part 4 consists of a methodology section where measurement tools, data description, used regression models and the target group are explained. In part 5 the results will be presented and interpreted for each of the models separately. Part 6 consists of a conclusion, followed by a discussion where limitations of this research are mentioned.
2. Literature review
Internet banking is the self-service technology of traditional banking, which allows the customers to use services, such as managing their bank account and transferring funds, without direct contact with an employee of the bank. Yang et al. (2009) state that the efficiency of the value chain is crucial for firm performance. Benefits of online banking are operating cost
minimisation and revenue maximisation. Internet banks can work at an expense ratio (operating cost divided by income from internet banking) of 15-20% compared to 50-60% for
the average bank (Booz, 1997). The usage of internet banking as a self-service technique is increasing. For instance, in Finland 39,8% of all retail banking transactions were made using internet in August 2000 (Mattila, M. et al 2003). In 2005 almost 2.6 million people in Finland used internet banking, which corresponds to 50% of the population. Considering globalisation ,the competition will be tougher because banks compete across their nation’s borders. Therefore, every country has to involve in internet banking. Adopting internet banking is almost impossible to circumvent, however, it is less easy to say why people keep using it while being aware of the danger of the cyber-threats (Yen and Gwinner, 2003).
2.1. Sense of security
Several papers, such as Akinci, S., Aksoy, S., & Atilgan, E. (2004) and Mattila, M., Karjaluoto, H. and Pento, T. (2003), are written about the adoption of internet banking. Researchers are aiming to find the factors that influenced the adoption of internet banking but in this paper the aim is to find what influences one of the factors. This factor is the sense of security of internet banking users. It is shown by Kolodinsky (2004) that in 2002 17% of the consumers adopted online banking in the United States although 91% of the households owned a bank account. Bebej (2003) estimated that this percentage would not exceed 30% of all households by 2007 . By conducting a survey Fisher (2007) found that only 23% of U.S
consumers use internet banking as a primary banking method. One of the possible influences could be that people don’t trust the system as it was relatively new in 2003 and hasn’t proven itself yet. So there is a difference in the behavior of internet banking customers and the adoption of the system between countries. The percentages of Kolodinsky(2004) and Fisher(2007), respectively 17% and 23%, are about the United States. Whereas the percentages from Finland found by Matilla M., et al (2003) indicate that in Finland internet banking is adopted relatively fast (39.8% of all retail banking transactions were made by using online banking in 2000).
Miyazaki and Fernandez (2001) found that a higher level of internet experience resulted in lower perceived risk and fewer concerns regarding system security and online fraud. They evaluated the level of internet experience from two perspectives: 1) The duration, meaning the elapsed time since a customer’s first internet access. 2) The frequency of internet usage which was measured by asking customers how much days per month they used web browsers such as Internet Explorer. They also noted that the mean expenditure of online purchases for consumers with a high sense of security was twice as low as the mean online purchase rate for consumers with a low to moderate sense of security.
Privacy and security aspects are mentioned in the research of Maenpaai et al (2008) as the general aspect of familiarity, defined as: ‘‘the number of product related experiences that have been accumulated by the consumer.’’ Familiarity, also referred to as user experience, increases trust and transparency due to cumulative successful transactions and usage. According to them this decreases insecurities because people feel more attached. Gefen (2000) mentioned that there is a difference between familiarity and trust. Familiarity is about understanding the current actions of other people or objects. Trust deals with the beliefs people have about other people and objects future actions. He states that the trust of the consumer increases when familiarity increases, and the consumer is more willing to shop online or do online transactions. Both trust and familiarity are measured by a Likert scale and information from websites such as how many times an individual visits the same website again. Speece and Rotchanakitumnuai (2003) introduce three barriers related to trust: Security, reliability of transactions and faith in the provider. With security being the main factor in their theoretical article because this incorporates everything from system security against hacks to the feeling that people could have while doing large money transactions without being able to see or hold the actual money. They interviewed several Thai corporate customers who mentioned that security is their main concern.
The adoption of internet banking is not only affected by a banks’ service quality, but also by their countries features (Levesque and McDougall, 1996). For example, a good quality and prestige bank such as the Royal Bank of Scotland does not necessarily experienced a faster adoption rate. With this fast adoption rate is meant how many months it takes customers
to go from traditional banking at the RBS to online banking at the RBS. The moment of conversion to internet banking is when the customer, registered as client, logs in to his online account the first time. Possible reasons for this are demographic profiles and the developing state of the country. Difference exist between countries regarding their banking and payment system. Yang, J et al. (2009) state that customers from the United States prefer credit card as a payment method. In contrast, Chinese customers prefer cash or debit card. Furthermore, the banking industry in the USA competes in all fields whereas Chinese banks are fully or partially government owned.
Liao and Cheung (2002) state that to find the perceived usefulness of e-banking,, measured by a 7 point Likert scale, one has to consider the expectations of the customers regarding the following attributes: security, transaction speed, accuracy, user-friendliness and convenience. The perceived usefulness can be interpreted as how useful e-banking is for customers based on their expectations of the attributes written above. The expectations of the customer are important because positive expectations lead to actions, in this case, using online banking. For example, if it can be shown that many customers doubt about the security of the online banking system, it is possible to reduce this doubt by signaling extra security measures. 2.2. Mobile usage
There are different platforms to access the internet such as by phone or by laptop. For instance, Chin et al. (2012) indicates thatpeople start tasks at the smartphone but are complete it at the computer. These platform switches occur due to differences in screen sizes or software, but they also believe that it is due to privacy and security concerns. With a smartphone people use public Wi-Fi more often than with their personal computer. However, many people are not fully aware of the dangers of public networks. Once you are on a public network, it is easy for a hacker to break into your system with a method called WireShark. Chin et al. (2012) suggest user education and improved interfaces to address misconceptions, such as failure to recognize the danger of public networks considering hacking, about Wi-Fi or other wireless networks. This education and improved interfaces make people more aware of the danger of public network and therefore their sense of security changes. The age of the participant and their concerns about privacy on mobile phones vs. laptops. Participants who are aged 48 years or older worried about privacy and security on their mobile phone more than on their laptops( Chin et al. (2012)
2.3. Age as influencer
Since internet banking is introduced in 1981 by the city bank of New York, the age of an individual has a possible influence on feelings of security and satisfaction from internet
banking because older people are more likely to experience stress from new technology( Elder et al.(1987)). In China,77% of internet users fall into the age group of 18 and 30 in 2001, and 86% of them have college degrees and above (Liao and Cheung ,2002). Older people are on average less familiar with the internet, mentioned by Chen and Persson(2010). Kim et al. (2005) state that the probability of people using internet banking decreases with age.
Moreover, Lassar, Manalis and Lassar (2005) state that there is a relation between the adoption of new technology and personal characteristics. The people who adopt new technology are usually younger and have higher levels of income and education. In the research of Laforet and Li (2005) it is noticeable that the ones who are interested in online banking are the younger people. The response rate in their research of the age between 25-34 is 57% and the response rate of the age 45-54 is only 6.3%. Age hs a general influence on computer usage besides online banking. Zeffane and Cheek's(1993) study in an telecommunications organization illustrated that age is negatively correlated with computer usage. Eriksson and Nilsson (2007) found no differences in age regarding perceived usefulness. However, they constructed two age groups (below 50 and above 50) so that their results conclude there is no difference between people below 50 and above 50 regarding how useful these groups find internet banking.
2.4. Control variables
Eriksson and Nilsson (2007) analyzed 1831 responses by letting people complete a survey. They came to the result that income levels are important to sense of security, as people with higher income actually use online banking more often according to them.
Kumbhar (2011) suggests that demographics of the customer are factors influencing internet banking usage. He shows that highly educated(University or higher), high income( above countries average income) and young( below 35 years old) individuals use internet banking more often. Monthly income and education levels indicate an individual’s cognitive abilities which also influences their sense of security because they are probably more aware of the issues of internet banking.
Another factor for consideration is the type of bank customers use. Specifically, if it is one of the big 3 banks ( ING, RABOBANK, ABN AMRO) as they have taken more security measures than smaller banks. The bank ING will also be taken into account by itself. Because of the recent scandals mentioned before it is interesting to analyze whether this had an impact on the sense of security of internet banking users.
The next factor is gender. Females are perceived to doubt more, they are more prone to stress when exploiting new technologies ( Elder et al. 1987). Harrison and Rainer (1992) found a relation between gender and computer skills, with males more likely to have better
computer skills. Furthermore, internet accessibility captures the platform and network of the customer. The use of a mobile phone is indicated to relate to a low sense of security( Chin et al. 2008).
3. Empirical review
Miyazaki and Fernandez (2001) conducted a survey amongst 161 people with ages from 15 to 75 years old. The mean purchase rate was 1.50 for consumers with a high sense of security and 3.13 for those with a low sense of security. The online purchase rate is measured by dividing the annual online purchases by the duration of internet experience. They found that security concerns are negatively associated with the online purchase rate. The coefficient they found is r= -.48 and is statistically significant at the level of 1%, showing that an increase in sense of security(less security concerns) is associated with a decrease in the online purchases rate.
Gefen (2000) found that trust and familiarity positively influence consumer’s willingness to shop online. Trust and familiarity both lower security concerns because of repeated successful interactions with the website. The coefficients from his regression model are r=0.53 and r=0.17 for trust and familiarity respectively. Both attributes are significant and are measured by an questionnaire via a Likert scale.
The differences in internet usage between countries are examined by Yang et al.(2009). For example, the usage of internet banking in terms of frequency differs substantially; in the USA 67.8% of the respondents uses internet banking once or more per week, whereas for China this percentage is 7.25%. Therefore it is important to analyze people from one country to find a relation between age and sense of security.
Being on a public network has consequences as (almost) any sensitive information can be read from your mobile phone by an apprentice hacker. Chin et al. (2012) found that overall, people are more willing to access their bank account on the laptop than on a mobile phone (p =0.0027). They also showed that 51% of the participants is concerned about security and privacy issues on the mobile phone versus 13% on the laptop. The p-value for the effect of age on the preference of laptop usage above mobile usage for internet banking is 0.011. People above 48 years old are 80% more probable to worry about internet usage on the mobile phone than younger people.
Kim et al.(2005) conducted a probit regression for internet banking usage. The predictor variable age has been divided in four groups. The first group consisted of ages between 35 and 50,the second between 50-65 and third above 65. The group with age below 35 was the reference group. Kim et al.(2005) found marginal effects of 0.0632 , 0.1337 , -0.2218 for groups one, two and three respectively. This means that older people a lower probability of using internet banking. Specifically, an individual older than 65 years has
22.18% lower probability of using internet banking compared to someone below 35 years old. Chen and Persson(2010) noted that older people are less familiar with internet because they use it less often. They showed that all young people (17-30 years) use the internet with an average of 400 minutes per week. But only 18% of the older people (60+) used the internet and at most 331 minutes. Kumbhar (2011) found that the ones who use internet banking the most are males(81%). Most of them (36%) are businessman and have a moderate to high income(54%).
Zeffane and Cheek's (1993) found significant effects of age on computer usage. In their research they found an average coefficient of r= -0.12, indicating that increases in age are associated with decreases in computer usage. They state unfamiliarity and low security feelings as possible reasons. The predictor variables for the usage of internet banking analyzed by Kumbhar(2011) are all significant. Education, income and age have p-values of p=0.027, p=0.022 and p= 0.02 respectively.
4. Methodology 4.1. Regression method
To analyze the dependent variable sense of security two ways are used. First, the respondent will be asked to give an indication (percentage) about how safe he or she feels about internet banking. These percentages are used to regress the sense of security of the respondent on the predictor variables age, male, income, education, frequency and bank size. The regression used for this is the ordinary least square regression. In the appendix the diagnostic test for residuals are presented. Showing normality and independency of the residuals, so that, ordinary least squares is assumed to be accurate.
To facilitate the effect of sense of security an regression to test the determinants(including sense of security) of the frequency of use of internet banking services will be done:
Second, a proxy considering mobile use is used. The regression used for the model about mobile usage is the logistic regression. The dependent variable is not continuous but binary,
therefore, the logistic regression is appropriate. Information about the logistic regression and its advantages over ordinary least squares can be found in the appendix.
The logistic model is: Pr(mobile)=
The logistic model will be analyzed by interpreting the marginal effects of each predictor variable on the dependent variable mobile usage.
4.2. Measuring sense of security:
Based on previous literature the most important factor of interest for the satisfaction of customers and adoption of internet banking, is the expectation customers have towards the security level. In this research the interest is not to find the determinants of adoption and satisfaction, but to find the determinants of the sense of security of customers. The aim is to conduct and analyze a model which displays the factors influencing the sense of security of internet banking customers. Expectations are that if the security concerns decrease there will be an increase in usage of internet banking. Therefore, it is important to know which factors affect the sense of security amongst internet bank users.
Sense of security is not an objective attribute because it has includes people’s own perceptions of security. The central aspect here is how secure the customer feels. Respondents are asked to give a percentage of how safe they feel while internet banking. Comparing respondents is hard because there is no base measurement. For example, one might over value his or her perception because this person doesn’t want to admit that he or she is insecure about internet banking. For this reason a second measurement tool is used as a proxy to security. Respondents are asked whether they dare to use their mobile phone for internet banking. This is an objective measure as the only answers are yes or no.
Table 1 shows the descriptive data of the response variables. The mean sense of security is 50.66%, showing that there is room for improvement. Whether people dare or don’t dare to use their mobile phone for internet banking is almost equal, 60% uses their mobile phone. The mean frequency is 5.56, meaning that on average people use internet banking 5.56 times per month.
Table 1 descriptive data response variables.
Variable descriptive
profile Number Percentage Mean SD Min Max
Sense of security <25 9 18% 50,66 22,48 5 95
(in percentage) 26-50 18 36%
51-74 15 30%
>75 8 16%
i. Mobile usage 0,6 0,495 0 1
for internet Yes =1 30 60%
banking No=0 20 40% Frequency(monthly) 5,56 3,051 1 15 < 4 19 38% 5-9 27 54% >10 4 8% 4.3. Target group:
Differences exist in the adoption and use of internet across countries. This research will consist of Dutch customers. In the Netherlands 81% of the population frequently used the internet weekly in 2011, whereas the European average is 57%(European Interactive Advertising Association, 2012). Furthermore, 70% of the Dutch population uses internet banking in 2012( CBS, 2013). Because in the Netherlands many households own a computer and use internet banking often it is relevant to conduct the research on Dutch people. The relation between age and sense of security can be expected to be measured more easily. The chance that older people actually use internet banking is greater.
4.4. Data collection and variables
To collect data for this research a survey is conducted. This survey will be filled in by 50 individuals. There were 23 people who completed the survey online. The other 27 respondents completed the survey orally. The information gathered from the survey is collected over a time of 2 weeks. This facilitates independency of the data because it eliminates time as a coherent factor.
The survey includes questions about individuals age, gender, education level, monthly income, monthly usage(frequency) and use of bank. Age, frequency and monthly income are continuous variables. Dummies are used for the other variables including Male (male=1), Education (University=1) and the size of the bank they use ( Size = 1 for large banks Abn Amro , ING or Rabobank ). Investigating ING is interesting because of the recent cyber-attacks on ING. Therefore, a dummy ING=1 is included. The survey questions are in the appendix.
4.4.1 Description of survey data. Table 2. Variable descriptive
profile Number Percentage Mean SD Min Max
i.Gender 0,64 0,4849 0 1 Male =1 32 64% Female=0 18 36% Age(years) 40,06 16,97 18 78 18-25 15 30% 26-35 11 22% 36-45 6 12% 46-55 9 18% 56-65 6 12% > 66 5 10% i. Education 0,54 0,503 0 1 University=1 27 54% Below University=0 23 46% Income(euro's) 2260,8 1771,43 290 8500 < 500 7 14% 501-1500 14 28% 1501-2500 9 18% 2501-3500 9 18% 3501-4500 6 12% 4501-5500 3 6% > 5501 2 4% Frequency(monthly) 5,56 3,051 1 15 < 4 19 38% 5-9 27 54% >10 4 8% i. Size Bank 0,72 0,454 0 1 big ABN,RABO,ING =1 36 72% small Other = 0 14 28% ING 0,38 0,49 0 1 ING =1 19 38% No ING =0 31 62%
The notation “i.” in front of some of the predictors stands for Indicator variable.
5. Results
Hierarchical results linear model on sense of security.
Table 3, regression table for sense of security. Showing the coefficient, (standard deviations)and *significance.
Sense of security
SoS (1) SoS (2) SoS (3) SoS (4) SoS (5) SoS (6) SoS (7) Robust
Age (in years) -1.01*** (.124) -.983*** (.127) -.599*** (.149) -.665*** (.145) -.623*** (.193) -.623*** (.195) -.623*** (.198) -.623*** (.178) Male 4.21 (4.48) 1.39 (3.97) .347 (3.82) .097 (3.93) .289 (4.07) .391 (4.14) .391 (4.62) Frequency (usage per month) 3.27*** (.839) 3.40*** (.805) 3.51*** (.874) 3.48*** (.892) 3.48*** (.903) 3.48*** (.788) Income (Monthly income in thousands of euro’s) 2.42** (1.01) 2.37** (1.07) 2.39** (1.08) 2.38*** (1.11) 2.38** (1.1) Education 1.64 (4.86) 1.36 (5.08) 1.38 (5.14) 1.38 (5.17) Size ING .949 (4.37) 1.25 (4.61) -.898 (3.94) 1.25 (4.19) -.898 (3.71) Constant 91.1*** (5.38) 87.3*** (6.68) 55.6*** (10.1) 52.7*** (9.68) 49.8*** (13.1) 49.2*** (13.5) 49.3*** (13.6) 49.3*** (13.7) N R2 Adjusted R2 50 .58 .57 50 .59 .57 50 .69 .67 50 .72 .69 50 .72 .69 50 .73 .7 50 .72 .68 50 .71 ***= 1%, **=5%, *=10% significance.
5.1. Interpretation results linear regression (sense of security and frequency of usage)
Table 2 shows that the null hypothesis of proposition 1 is rejected. There is a (negative) linear relationship between age and sense of security with a coefficient that significantly(pvalue<1%) differs from zero. A one year increase in age means that the sense of security of a person decreases with 0.623%, suggesting that older people feel less secure while internet banking. The results of Eriksson, K., & Nilsson, D. (2007), who found no significant different in age regarding usefulness of internet banking, do not directly contradict the results of this paper because the response variables differ. However, Zeffane and Cheek (1993) found corresponding results. They found that older people feel less secure on the internet overall. Kim et al.(2005) showed that older people use internet banking less often. A regression of age on frequency confirmed this, r= -.063 (see table 4). Taken these two researches into consideration it can be argued that older people use internet banking less often due to that they feel less secure about it compared to young people, this is confirmed by the results in table 3. The gender of the respondent seems to be insignificant. This is in contrast with Elder et al.(1987) who argued that females use new technology less because they feel more insecure about it than males. A reason for this might be that internet banking is not a relatively new technology anymore, so that males and females are both evenly used to it. Gender has no significant effect on the frequency of usage neither.
Frequency of usage per month is a significant predictor of sense of security. A one-unit increase in frequency per month is associated with a 3.48% increase in sense of security. Thus, people who use internet banking often have a high sense of security, in other words, they feel safe doing it. Miyazaki and Fernandez (2001) showed that the other way around holds true. People who feel 1% more safe during internet banking tend to use it .075 more frequent ( see table 4). This results are interesting for banks because if they manage to increase sense of security by for example 50% that as a results people will use internet banking 3.75 times more per month.
Monthly income of an individual has an significant effect people sense of security while internet banking. A thousand euro increase in monthly income is associated with a 2.38% increase in sense of security. This means that people who have higher income feel more safe with online banking. A reason for this might be that people with high incomes also use internet banking more often. Businessman use e-banking often and have relatively high incomes( Kumbhar, 2011). However, a simple test of income on frequency resulted in insignificant results(table 4).
Education levels are not significant. There is no significant effect of being of university degree or not on the sense of security while online banking. Kumbhar(2011) found a significant relation between education and internet banking usage, however, this does not imply that higher educated people use internet banking more often than low educated people because of security concerns. University graduates use internet banking 1.73 times less per month than graduates from lower levels of education. There is no significant effect of the size of the bank where people use
internet banking. Even more interesting, despite the recent cyber-attacks and the announcement of Hans Hagenaar (Director at ING) about sharing purchase information to third parties, there is no significant effect of the variable ING. Thus, sense of security is not affected by whether someone banks at ING or not.
Table 4. Regression of age, male, income and education on frequency of internet banking usage.
Frequency per month (1) (2) (3) (4) (5) (6) (7) Robust Age (in years) -.122*** (.019) -.117*** (.019) -.043* (.026) -.026 (.027) -.064** (.030) -.063** (.03) -.063** (.03) -.063** (.026) Male .859 (.677) .541 (.599) .612 (.591) .788 (.567) .861 (.583) .869 (.593) .869 (.556) Sense of security .076*** (.019) .083*** (.02) .076*** (.019) .075*** (.019) .075*** (.019) .075*** (.018) Income (Monthly income in thousands of euro’s) -.268* (.132) -.191 (.164) -.186 (.166) -.187 (.171) -.187 (.185) Education -1.62*** (.675) -1.73*** (.699) -1.73*** (.708) -1.73** (.753) Size ING .401 (.639) .424 (.675) -.073 (.578) .424 (.557) -.073 (.574) Constant 10.46*** (.826) 9.70*** (1.02) 3.08 (1.92) 2.59 (1.9) 5.06*** (2.08) 4.80** (2.14) 4.82** (2.17) 4.82** (1.78) N R2 Adjusted R2 50 .46 .45 50 .48 .46 50 .61 .58 50 .63 .60 50 .67 .64 50 .68 .64 50 .68 .63 50 .68 ***= 1%, **=5%, *=10% significance.
5.2. Logistic regression on mobile usage.
Table 5, regression table for Mobile usage. Showing marginal effects , logistic coefficients (standard deviations) and significance*.
Mobile(0,1) Marginal effect Coefficients logit regression Age (in years) -.0312*** (.0066) -.0225*** (.0041) Male .335*** (.128) .217* (.112) Income (thousands of euro’s per month) .0135** (.048) .049 (.031) Education -.398*** (.173) -.224* (.134) Size .219* (.131) .124 (.248) *=10% **=5% ***=1% significance
The null hypothesis of proposition 2 has been rejected. The marginal effects for the variables age, male, income, education and size are all significant. A one year increase in age means a probability decrease of 3.12%. This is conformable with the study of Zeffane and Cheek (1993) who presented that older people use the internet significantly less, so that, they also use the internet of their mobile phone less. It is also confirmed by the results of Chin et al.(2012) who found that people of older age prefer their laptop over their mobile phone for internet banking. Arguments they give are differences in screens size, interface and difficulties with touchscreens. The probability that an individual will use his or her mobile phone for internet banking purposes is 33.5% bigger for males compared to females. This results correspond to the study of Elder et al. (1987) who found that females use (new) technologies less often than males due to that they doubt more and feel less secure about it. Every one euro increase in monthly incomes means a probability increase of 1.35%. This is agreement with Kumbar’s(2011) findings who showed that most mobile internet users are businessman with moderate to high income. Individuals who have done or are doing a study at an university have a 39.8% lower
probability than people who did not study at a university.
by Chin et al. (2012). The marginal effect of size shows that people who bank at a large bank are 21.9% more willing to use their mobile phone for internet banking.
6. Conclusion, limitations and directions for future research
In this thesis information has been gathered by conducting a survey amongst 50 internet banking users. The aim is to find an effect of age on the sense of security of internet banking users. In the survey people answer a percentage as an indication of sense of security. Analyzes been done by conducting a linear regression model regarding sense of security, a linear regression regarding the frequency of usage and a logistic regression regarding mobile usage. Results of the linear regression model shows that age has a significant negative influence of -.623 on the sense of security of internet banking users in the Netherlands. A 1 year increase in age means a 0.623% decrease in sense of security. Income per month is significantly positively related to sense of security. A thousand euro increase in monthly income is associated with a 2.38% increase in sense of security. This reinforces the findings of Kumbhar (2011). Frequency of usage has a positive significant influence of on the sense of security, a one-unit increase in frequency per month is associated with an 3.48% increase in sense of security. The other way around holds true as shown by table 4, an 1% increase in sense of security increases frequency of use by 0.075. This reinforces Maenpaai et al.’s (2007) research about familiarity. The variables male, education, bank size and ING are insignificant. However, it is significantly shown that males are 33.5% more likely to use their mobile phone for internet banking purposes and that people with university degrees are 39.8% less likely to use their mobile phone.
7. Discussion and limitations.
Information has been gathered by a survey amongst 50 people and analyzes has been done by using regressions. Interpretations of the results have to be taken with caution because the response variable sense of security is an indicator of peoples beliefs. People have to fill in a percentage regarding their feeling of security without having a base outcome to compare with. Tiebling et al., 2006 argues that questions regarding people beliefs can results in a bias as people tend to overvalue or undervalue their perceptions. In other words, ex ante valuations often exceed ex post utilization (Tiebling et al., 2006).
Results of this research are only useful regarding Dutch customers. More than 90% of the respondents of the survey live in the provinces Noord-Holland, Zuid-Holland or Utrecht. Future researchers with more time and resources are suggested to collect respondents across all provinces of the Netherlands. Furthermore, this research consisted of 50 respondents, future research is suggested to find more respondents(preferably 800 or more to construct accurate mat size tests and other statistical tests) if there is time for it. Also, with more observations it is possible to perform an accurate cluster-analysis. With such an analysis it is possible to find and arrange groups that for example, use internet banking often and focus most on interface convenience. Cluster-analysis makes it possible to
find if these groups that are to be treated the same way. Groups with people of for exampl a particular age, education and gender. With such information the marketing aspect of the research is enhanced because then banks can target each group individually on the aspect the group values the most. Such an analysis takes more time and observations, therefore, it is not possible to examine it in this article.
8. Appendix
Appendix table 1: The correlation matrix for all variables.
Correlation ss age male income education size mob frequency
ss 1.000 age -0.7614 1.000 male 0.2506 -0.2155 1.000 income -0.0683 0.2977 0.0369 1.000 education 0.4096 -0.6033 0.2274 -0.0627 1.000 size 0.2326 -0.2920 -0.0965 -0.1562 0.3182 1.000 mob 0.6661 -0.6023 0.3748 -0.0015 0.3529 0.1913 1.000 frequency 0.7605 -0.6807 0.2770 -0.2372 0.2109 0.2041 0.5870 1.000 Appendix table 2:
Coefficients (log odds ratio) of logit regression.
Mobile(0,1) Mobile Mobile Mobile Mobile Mobile Robust Test Age (in years) -.1138*** (0.0298) -.1141*** (0.031) -.1037*** (0.0306) -.2442*** (0.0892) -.3015*** (0.1084) -.3015*** (0.101) Income (in euro’s) 0.0713 (0.2) -.1 (0.2) 0.37 (0.35) .6 (.42) 0.58 (0.52) Male 1.4008* (0.8398) 1.9489** (0.9819) 2.6871** (1.1842) 2.687* (1.517) Education -4.2336** (2.1069) -6.0929** (2.6557) -6.093** (2.607) Size 1.912 (1.2256) 1.192 (1.2622) Constant 5.214*** (1.3372) 5.2039*** (1.3643) 4.1612*** (1.4131) 10.9230*** (4.0320) 12.0613*** (4.5466) 12.06133*** (3.9641) Pseudo R2 0.3754 0.3754 0.4186 0.5088 .5493 0.5493 Chi2 p-value 0.0 0.0 0.0 0.0 0.0 0.0178 *=10% **=5% ***=1% significance
7.1 Residuals diagnostic tests: Normality, heteroscedasticity, multicollinearity.
Tests for normality.
Null hypothesis: The residuals are normally distributed.
Based on both P-values and the joint P-value the null hypothesis that the residuals are normally distributed can’t be rejected. For further information about the distribution consult appendix graph 1. Here it can be seen that the residuals are somewhat normally distributed. The robustness test in table 3 is used to correct for possible misinterpretations about the normality.
P-value Skewness P-value Kurtosis Joint adjusted Chi2 Joint P-value
0.4649 0.5012 1.03 0.5986
Appendix graph 1a and b: residuals are normally distributed.
Test for heteroscedasticity:
Breush-Pagan/Cook-Weisberg test: Null hypothesis: Constant variance(homoscedastic)
Chi2 0.36 P-value 0.548 0 .0 1 .0 2 .0 3 .0 4 D e n si ty -20 -10 0 10 20 Residuals 0 .0 1 .0 2 .0 3 .0 4 D e n si ty -40 -20 0 20 40 Residuals
Kernel density estimate Normal density
kernel = epanechnikov, bandwidth = 4.7290
The null hypothesis has been rejected, residuals are not constant. They might be heteroskedastic. Therefore, robustness test is important because this deals with heteroskedasticity. As shown in
appendix graph 2. Appendix graph 2:
Test for multicollinearity:
When collinearity increases the regression coefficients become unstable, the standard errors inflate. A variance inflation factor (vif) test gives insights to whether there is multicollinearity.
Variable VIF 1/VIF
age 3.38 0.295692 frequency 2.28 0.437789 education 2.02 0.495803 big3 1.32 0.759217 male 1.21 0.825325 incometh 1.15 0.873235 ING 1.12 0.892557 Mean VIF 1.78
A vif > 10 indicates multicollinearity. In this regression there is no significant multicollinearity.
-2 0 -1 0 0 10 20 30 R esi du al s 0 20 40 60 80 100 Fitted values
7.2 Logistic regression.
The basic principle of the logistic regression is that it is based on finding the probability of membership for each category. In this research, the focus is on what the probability is for a Low, Moderate or High sense of security. It compares the probability of being in n-1 categories compared to a baseline. For instance, the baseline is Moderate(2). The model estimates the effect of for instance Age to being in Low or High compared to Moderate.
With a logistic regression one calculates the odds ratio, which is the probability of the event occurring divided by the probability of the event not occurring. Probability is defined as the number of times the event occurs (mob=1) divided by the number of times it could occur. Probabilities and odds are often used as substitutes in normal vocabulary, in statistics however, odds are defined differently. The Odds of an event happening is the probability that the event occurs divided by the probability that it does not occur. An example for this research would be: 75% of men use their mobile phone for IB and 60% of women use their mobile phone for online banking. The odds for men are .75/.25 =3 and for women .6/.4=1.5. The odds ratio will be: 3/1.5=2, meaning that the odds are 2 to 1 that a man will use his cellphone for banking compared to a female. The last term used in logistics is the log odds, also known as logit. The coefficients resulting from a logit regression are given in units of log odds. A coefficient indicates the amount of change expected in the log odds if there is a one unit change in the predictor variable, ceteris paribus. A logit is defined as the log base e (log) of the odds. : logit(p) = log(odds) = log(p/q) , with q=1-p
The estimated coefficients are interpreted in a different way than with OLS. Instead of the coefficient being the rate of change in the response variable as the predictor variable changes, it is now interpreted as the rate of change in log odds of the response variable as the predictor variable changes. The calculated coefficients are the same as the log odds and can be interpreted as: e^(β0 + β1 AGE). The odds ratio can be hard to interpret due to its complexity especially for continuous variables, such as age. Factor variables can be compared with odds ratio’s because it is one outcome vs another single outcome. Marginal effects are helpful to compute to help interpret the results of continuous variables because this measures the effect of a variable on the response variable per unit or particular value.
The logistic regression assesses the model by using the log-likelihood statistic. This is an indicator of how much unexplained information there is after the model has been fitted. So, large values indicate poorly fitted models.
The predictors can be assessed by using the Wald chi2 Statistic. Chi2= b/SE which is similar to the t-statistic in linear regression. The null hypotheses b=0 is tested. The coefficients are presented in log odds as explained above. To calculate the odds ratio from the log odds
coefficient one has to reverse the log by using exp(coefficient). This indicates the change in odds resulting from a unit change in the predictor. Ratio > 1: When the predictor increases, the probability of event outcome occurring increases. Ratio<1: When the predictor increases, the probability of the event outcome decreases.
The minimum regression in logistic terms is:
In logistic regression probabilities are used to evaluate the response variable, so we don't need an error term in the model. Probabilities are themselves ways of dealing with randomness and imprecision so there is no error term necessary. For instance, the probability of success is 60% this means that the probability of failure is 40%. The main relationship of interest is the effect of age on the sense of security. Other variables are used as control variables; they also have a possible relationship with the dependent variable.
Survey questions:
Geslacht: Man Vrouw
Leeftijd: Wat is uw leeftijd?
Frequency: Hoevaak gebruikt u internet bankieren per maand? Inkomen Wat is ongeveer uw gemiddelde maand inkomen?
Education: Wat is uw hoogst genoten opleidingsniveau? (inclusief in-progress)
Size/welke bank: Bij welke bank doet u de meeste online transacties? ( ING, RABOBANK, ABN AMRO, OTHER)
Sense of Security: Gebruikt u wel eens voor internet bankieren uw mobiele telefoon?
In hoeverre vind u dat u gevoelig bent voor de veiligheid van internet bankieren? Geef a.u.b. een percentage.
7.3 Data table.
Age in years, income in euro’s, frequency per month and ss_proc in percentages.
ID age Male(=1) income education Size(Big=1) Mob(yes=1) ING(yes=1) frequency ss_proc
1 34 1 3000 1 1 1 1 8 55 2 39 1 2500 1 1 1 0 9 60 3 56 0 750 0 1 0 0 3 25 4 21 1 550 1 1 1 1 11 85 5 20 1 500 0 0 1 0 9 80 6 45 0 1500 0 1 1 0 10 60 7 18 0 300 1 1 1 0 7 90 8 28 1 1800 1 1 1 0 4 70 9 65 1 1100 0 0 0 0 2 25 10 55 0 2900 0 0 0 0 5 33 11 36 1 3900 1 0 0 1 4 45 12 22 1 500 1 0 1 1 8 70 13 24 1 700 1 1 1 0 8 75 14 27 1 750 1 1 0 0 8 80 15 47 0 3500 0 1 1 0 5 65 16 55 1 2000 0 0 0 0 5 35 17 34 0 2600 1 1 0 1 4 60 18 31 0 2800 1 1 0 0 5 40 19 49 1 3200 0 0 1 0 5 45 20 50 0 4000 0 0 0 0 5 25 21 57 1 1800 0 1 0 1 2 30 22 49 0 3000 0 0 0 0 2 40 23 66 0 1900 0 0 0 0 1 5 24 27 1 2000 1 1 1 1 12 65 25 53 1 5400 1 1 0 0 3 25 26 26 1 800 0 1 1 0 9 45 27 47 0 6300 1 1 1 0 4 75 28 33 1 8500 0 1 1 1 5 70 29 18 0 290 0 1 1 1 9 60 30 19 1 400 0 1 1 0 15 95 31 22 0 500 1 1 1 1 4 45 32 18 1 500 1 1 1 0 5 45 33 28 1 900 1 1 1 0 5 50 34 40 1 4000 1 1 1 0 7 50 35 78 1 1400 0 0 0 0 1 5 36 56 1 5000 0 0 1 0 5 70 37 37 1 4800 1 0 1 0 7 65 38 45 0 2300 1 1 0 0 2 30 39 22 1 700 1 1 1 1 4 40 40 27 1 2800 1 1 1 1 7 70 41 66 1 2700 0 1 0 1 4 45 42 22 0 550 1 1 1 1 9 80 43 23 1 650 1 1 1 0 5 65 44 62 0 4350 0 1 0 0 2 15 45 61 1 3550 0 1 1 1 5 35 46 73 0 800 0 1 0 1 1 5 47 24 1 1900 1 1 1 1 8 65 48 33 1 3600 1 0 1 0 5 50 49 47 1 2000 1 1 0 1 3 45 50 68 0 800 0 1 0 1 2 25
References.
Akinci, S., Aksoy, S., & Atilgan, E. (2004). Adoption of internet banking among sophisticated consumer segments in an advanced developing country.International Journal of Bank Marketing, 22(3), 212-232.
Booz, Allen & Hamilton (1997) ``Internet banking:a global study of potential'', Booz, Allen &Hamilton Inc., New York, NY
Chen, Y., & Persson, A. (2002). Internet use among young and older adults: Relation to psychological well-being. Educational Gerontology, 28(9), 731-744.
Chin, E., Felt, A. P., Sekar, V., & Wagner, D. (2012, July). Measuring user confidence in smartphone security and privacy. In Proceedings of the Eighth Symposium on Usable Privacy and Security (p. 1). ACM.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of two theoretical models. Management science, 35(8), 982-1003.
Elder, V.B., Gardner, E.P. and Ruth, S.R. (1987), ``Gender and age in technostress: effects on white-collar productivity'', Government Finance Review, Vol. 3 No. 6, pp. 17-21.
Eriksson, K., & Nilsson, D. (2007). Determinants of the continued use of self-service technology: The case of Internet banking. Technovation, 27(4), 159-167.
Fisher, M. 2007. Why banks still grow branches. Washington Post(September 18) B3.
Gefen, D. (2000). E-commerce: the role of familiarity and trust. Omega, 28(6), 725-737.
Howcroft, B., Hamilton, R., & Hewer, P. (2002). Consumer attitude and the usage and adoption of home-based banking in the United Kingdom. International journal of bank marketing, 20(3), 111-121.
Kim, B. M., Widdows, R., & Yilmazer, T. (2005). The determinants of consumers’ adoption of Internet banking. Unpublished paper.
Kolodinsky, J. M., Hogarth, J. M., & Hilgert, M. A. (2004). The adoption of electronic banking technologies by US consumers. International Journal of Bank Marketing, 22(4), 238-259.
Kumbhar, D. V. (2011). Service Quality Perception and Customers' Satisfaction in Internet Banking Service: A Case Study of Public and Private Sector Banks. Cyber Literature , Vol. 4 (2), 21-30
Laforet, S. and Li, X.Y. (2005) ‘Consumers’ attitudes towards online and mobile banking in China’, The International Journal of Bank Marketing, Vol. 23, Nos. 4–5, pp.362–381.
Lassar, W. M., Manolis, C., & Lassar, S. S. (2005). The relationship between consumer innovativeness, personal characteristics, and online banking adoption. International Journal of Bank Marketing, 23(2), 176-199.
Levesque, T. and McDougall, G.H. (1996) ‘Determinants of customer satisfaction in retail banking’, International Journal of Bank Marketing, Vol. 14, No. 7, pp.12–20
Liao, Z.Q. and Cheung, M.T. (2002) ‘Internet-based e-banking and consumer attitudes: an empirical study’, Information and Management, Vol. 39, No. 4, pp.283–292
Mattila, M., Karjaluoto, H. and Pento, T. (2003), “Internet banking adoption among mature customers: early majority or laggards?”, Journal of Services Marketing, Vol. 17 No. 5, pp. 514-28.
Maenpaai K., Kale, H. K., Kuusela, H., and Mesiranta, N., (2008), Consumer perceptions of internet banking in Finland: The moderating role of familiarity, Journal of Retailing and Consumer Services, 15(4) pp. 266-276.
Miyazaki, A. D., & Fernandez, A. (2001). Consumer perceptions of privacy and security risks for online shopping. Journal of Consumer Affairs, 35(1), 27-44.
Reichheld, F.F., Sasser Jr., W.E., 1990. Zero defections: quality comes to services. Harvard Business Review 68 (5), 105–111.
Stock, J.H., Watson M.M., 2011. Introduction to Econometrics, 3rd edition. Pearson Education, pp.434-460.
Tiebing Shi, Jay M. Handelman, Shenghui (Jerry) Zhao, and Robert Meyer|Jin Han (2006) ,"Overvaluation Bias in the Valuation and Utilization of New-Product Attributes", in NA - Advances in Consumer Research Volume 33, eds. Connie Pechmann and Linda Price, Duluth, MN : Association for Consumer Research, Pages: 244-244.
Yang, J., Cheng, L. and Luo, X. (2009) ‘A comparative study on e-banking services between China and USA’, Int. J. Electronic Finance, Vol. 3, No. 3, pp.235–252.
Yen, H.J.R., Gwinner, K.P., 2003. Internet retail customer loyalty: the mediating role of relational benefits. International Journal of Service Industry Management 14 (5), 483–50
Zeffane, R. and Cheek, B. (1993), ``Profiles and correlates of computer usage: a study of the Australian telecommunications industry'', Computers in Industry, Vol. 22, pp. 53-69.
Zhao, H. (2002) ‘Rapid internet development in China: a discussion of opportunities and constraints on future growth’, Thunderbird International Business Review, Vol. 44, No. 1, pp.119–138.
