Investigating Different Levels of Mobile Banking Usage: A Mixed-Methods Study in China

M-banking usage has attracted great attention from literature in recent years. The goal of this study is to provide new insights into m-banking usage by investigating the motivations for m-banking usage and how these motivations can drive different levels of m-banking usage respectively. A quantitative dominant mixed-methods design is applied in this study. First, a qualitative study through interviews is conducted to identify the motivations for m-banking usage. Five m-banking motivations are identified from the qualitative study, namely convenience, cost saving, information gaining, word-of-mouth and decision-making quality. Second, a quantitative study through survey is conducted to test the effects of motivations on different levels of m-banking usage. In general, results indicate that different levels of m-banking usage could be driven by different m-banking motivations.


I. INTRODUCTION
M-banking has been widely adopted by banking industry in recent years. Most of the popular commercial banks in China have provided m-banking services to enhance customer value and relationship. A report 1 released by the China Financial Certification Authority (2018) shows that the ratio of m-banking user has come up to 57% in China, with annual growth of 6%. However, the growth of m-banking user is slowing down, and the m-banking market is entering the mature stage. Thus, one main challenge for m-banking industry is how to improve the level of m-banking usage of existing users, rather than the adoption of new users. Hence, more attention should be paid to the levels of m-banking usage, which has been ignored in the literature.
Although m-banking usage has recently attracted great attention from literature, existing literatures treat m-banking usage as homogenous overall usage (the overall behavior of employing the m-banking in completing banking tasks) and do not consider different levels of m-banking usage [1]- [3].
The associate editor coordinating the review of this manuscript and approving it for publication was Claudio Agostino Ardagna . 1 https://zhuanti.cebnet.com.cn/upload/report1.pdf Examining the overall m-banking usage is significant, especially when m-banking is in its infancy. However, it could not uncover the reasons why customers would make use of m-banking at different levels, which should be a crucial issue as m-banking being mature and widely accepted. Nevertheless, research on the levels of m-banking usage remains rare. Hence, investigating different levels of m-banking usage could provide both theoretical implications for m-banking usage research and practical implications for m-banking industry. Different from prior research, this study focuses on the levels of m-banking usage-the degree of which a user makes full use of the available m-banking service, which has been seldom investigated. Uses and Gratifications (U&G) theory has been widely employed to identify a range of motivations driving different usages in various web/mobile applications [4], [5]. From the perspective of U&G theory, user's motivations are very fundamental for understanding m-banking usage, and more importantly, the motivations that drive the different levels of m-banking usage could differ. Although literatures have investigated the m-banking usage from many theoretical perspectives, the motivations for m-banking usage have been seldom studied. Especially, our understanding on the roles of m-banking motivations in different levels of m-banking usage is insufficient. To sum up, our goal is to address the following questions: Q1: What motivations drive m-banking usage? Q2: How could various m-banking motivations drive different levels of m-banking usage; and further, what are the priorities of those m-banking motivations on different levels of m-banking usage?
To address these questions, we applied a quantitative dominant mixed-methods design which is useful for investigating both exploratory and confirmatory research questions. First, considering the exploratory question in Q1, a qualitative analysis by interviews was used to identify m-banking user's basic motivations (Study 1). Specifically, 32 interviews were performed to collect qualitative data; and a data coding process is conducted to identify the m-banking motivations. Second, to confirm the relationships between m-banking motivations and different levels of m-banking usage stated in Q2, we used quantitative analysis by a survey (Study 2). Specifically, we proposed a research model combing the U&G theory and findings from Study 1; after that, we collected quantitative data from an online survey and applied PLS-SEM to analyze the effects of m-banking motivations on different levels of m-banking usage. Further on, for post-hoc analysis, we also tested the effects of the motivations on overall usage, and then made a comparison between the motivations predicting overall usage and those predicting specific levels of m-banking usage. This comparison can provide evidence for the validity of disaggregating of different levels of m-banking usage.
Our findings can contribute to m-banking literature in several aspects. First, this study is one of the first to investigate different levels of m-banking usage. Second, this study can provide knowledge on m-banking usage by identifying motivations for m-banking usage through a qualitative study. Especially, this study reveals that decision-making quality as a m-banking usage motivation which has been ignored in the literature. Third, this study draws on U&G theory to examine how m-banking motivations could drive m-banking usage (including different levels of usage and overall usage). These findings suggest that disaggregating of different levels of m-banking usage might provide insights into how users make use of m-banking.

II. LITERATURE REVIEW AND THEORETICAL FOUNDATION A. UNDERSTANDING LEVELS OF M-BANKING USAGE
M-banking enables its users to get access to various bank services via a mobile device. Thus, m-banking can be considered as a collection of financial services used to conduct various banking tasks to meet different needs. The most common bank services available in m-banking include: account enquiry, account transfer, payment, credit card business and buying financial products. According to the actual use of specific service, there are different levels of m-banking usage.
This article defines a ''level'' as the degree of which a user makes full use of the available banking services, which means that the relatively high level should be based on the relatively low level. For example, we considered ''account transfer'' is a higher level of usage than ''account enquiry'', because users making use of ''account transfer'' is supposed to have made use of ''account enquiry'' in m-banking. According to the 2018 China E-bank Investigation Report, m-banking usage could be ordered into three levels, i.e. usage for account enquiry (UE), usage for account transfers (UT) and usage for investment (UI): UE refers to using m-banking to enquire account information such as balance and transfer record. UE is the lowest level, which only involves information flow but not fund flow.
UT refers to using m-banking to transfer money between different accounts. UT is the second level, which involves information flow and fund flow. UT can be implemented in the forms of paying bills, account transfer, credit card repayment, etc.
UI refers to using m-banking to conduct investment behavior, such as buying stocks and funds. UI is the highest level, which includes buying financial and investment products.  Table 1. In general, UTAUT2 and D&M IS Success Model are the most common theories for explaining m-banking usage. Based on the D&M IS Success Model, researches find that information quality, system quality and satisfaction can positively affect m-banking usage, but service quality cannot [6]. Literature also indicates the significant roles of the main construct of UTAUT2 in m-banking usage [1], [7], [8]. Another effective theory for explaining m-banking usage is TTF model which emphasizes the effect of the fit between task characteristics and technology characteristics on m-banking usage [9].
This study argues that m-banking usage is still need to be further studied for two aspects. First, prior researches are based on various typical IS utility theories, but seldom research has investigated the basic motivations for m-banking usage. Second, literatures treat m-banking usage as one common variable and do not differentiate different levels of m-banking usage; however, users can use m-banking to achieve various banking services and satisfy different needs. Hence, treating m-banking usage as one common variable might obscure our understanding about m-banking usage.

C. USES AND GRATIFICATIONS THEORY
U&G theory focuses on ''what and how individuals do with media'' rather than what media can do to individuals. U&G theory, a method to understand why and how individuals actively select specific media resources to meet their specific needs, is found to be effective in explaining individuals' usage in various web/mobile applications [4], [5], [12], [13]. U&G theory can provide insights into the underlying psychological mechanism of individual behaviors in using web applications by identifying individuals' specific needs from specific media VOLUME 8, 2020 services [4]. In line with U&G theory, this article argues that users make use of m-banking with various motivations, which in turn can play significant roles on their m-banking usage; what's more, different motivations would have different effects on different levels of usage in m-banking.

III. RESEARCH DESIGN
Mixing quantitative and qualitative research can offer more advantages than a single method research in terms of addressing simultaneously both exploratory and confirmatory research questions; one of the dominant and appropriate purpose for using mixed-methods design is developmental [14], [15]. The level of m-banking usage is a research issue that has been seldom investigated. Besides, the essential motivations for m-banking usage has been ignored. Taken together, this study belongs to developmental research that addressing both exploratory and confirmatory research questions. Therefore, we argue a mix-method design is appropriate for this study. Following mixed-methods design of Hua et al. [15], this study uses interview for qualitative research and survey for quantitative studies. As shown in Section. 4, study 1 identifies five motivations for m-banking usage through 32 interviews. Based on those five motivations, study 2 proposes the research model from the perspective of U&G theory, and conducts a survey to test it.

IV. STUDY 1: QUALITATIVE RESEARCH A. PARTICIPANTS AND DATA COLLECTION
Participants for interviews are m-banking users in China. Following French et al. [16], we use a convenient sample and networking to invite participants. All the participants have received university or higher education. Participants are volunteer, and they can get a gift after finishing the interviews. We invite 35 participants for interviews, and 3 participants could not finish the interviews because of time conflict. Finally, we have 32 participants for interviews.
The interview outline includes two sections. Section 1 is for asking basic information about m-banking experience, education, etc. Section 2 is for asking motivations for mbanking usage. Three questions are used in section 2 to identify participants' motivations for m-banking usage. The first question is: ''why do you use m-banking''. The second question is: ''Please tell me the benefits of using m-banking''. The last question is: ''what are advantages of m-banking over other banking channels''. The original interviews are conducted in China. Every interview lasts 15-20 minutes. Instead of verbal response, participates are required to response by written answer, which could better ensure accuracy of responses [16].

B. DATA CODING AND RESULTS
The data coding process is conducted by two researchers. First, two researchers complete the data coding independently. After that, they compare their coding with each other, and discuss whether they agree or disagree. Only the coding that are agreed by both researchers are retained. After discussion, two researchers reach a consensus and identify eight motivations for m-banking usage. Table 2 shows the eight motivations identified from data coding. Some sample quotations are presented in Appendix A.
The coding with less participants could be considered as insignificant and removed [16]. In this current study, we consider the coding with less than three participants (appropriately 10%) as insignificant. Two motivations for m-banking usage are removed.
Among the six significant motivations, convenience is the most frequently mentioned motivation. Convenience refers to the ability to interact with banks and access banking services anywhere and anytime, allowing users to access banking service without the constraints of time or space. Mobile devices are more convenient and accessible than other media platforms [13].
The second most significant motivation is cost saving. Cost saving reflects users make use of m-banking service in order to save cost for service. Price value is a main factor for actual use in m-banking [17]. For example, in China, many banks today provide free transfer service in the m-banking platform to attract users. Hence, some users use m-banking because it can help them save their cost for bank services.
The third motivation is information gaining. Information gaining is concerned with gaining accurate, relevant, upto-date and complete information. Information quality is a strong positive factor on attitude and behavior toward mbanking [9], which indicates that one basic motivation for using m-banking is gaining information.
The fourth motivation is word-of-mouth, indicating that users sometimes make use of m-banking because others (such as family or friend) advise them to do so. In the online context, word-of-mouth is useful for advertising a product or service. Mehrad and Mohammadi [18] find that word-of-mouth can significantly drive user's attitude and intention toward mbanking.
Diversified service is the fifth motivation derived from data coding. Diversified service is identified by the ability to access a variety of financial services via using m-banking. With development of m-banking, the services provided by m-banking are getting increasingly diversified. However, we find that most of the examples of diversified service are related to convenience. Considering the concept of diversified service, it can be inferred that diversified service should be a dimension of convenience. In order words, users who are motivated by diversified service is primarily motivated by convenience. For example, a participant provides the answer as ''It's very convenient to achieve various banking services via m-banking''. It can be argued that diversified service is a sub-motivation under convenience rather than a separate motivation. Therefore, diversified service will not be examined in Study 2.
The last significant motivation identified is decisionmaking quality. Decision-making quality is defined as ''a user's degree of confidence in a purchase decision'' [19]. Decision-making quality can play crucial roles in individual's behavior in online environment [20]. M-banking can be used to buy some financial products such as stocks and funds, in where users have demand for improving decision making quality. In conclusion, in regards to research question 1, five motivations for m-banking usage are finally identified from Study 1, namely convenience, cost saving, information gaining, word-of-mouth and decision-making quality.  different motivations. Also, it's worth noting that the motivations for different levels of usage are not necessarily exclusive, as indicated by prior researches [4][5]. That is, the motivations for different levels of usage could overlap. Consequently, based on the essence of various motivations, we propose that four motivations (i.e. convenience, cost saving, information gaining, word-of-mouth) can lead to all the three levels of m-banking usage; while decision-making quality might only drive UI.

1) EFFECTS OF CONVENIENCE
M-banking is widely considered as a more convenient channel to conduct banking practice [21]. One of the greatest advantages of using m-banking is convenience by allowing users to access financial services anywhere and anytime [22]. Users connected to technologies that are convenient [1]. Hence, users would prefer to make use of m-banking rather than other banking channel to conduct various banking practice including UE, UT and UI. Hence, we propose: H1a: Convenience positively affects UE. H1b: Convenience positively affects UT. H1c: Convenience positively affects UI.

2) EFFECTS OF COST SAVING
Cost saving is related to the construct of price value which reflects users' cognitive trade-off between the perceived benefits of using m-banking and the monetary cost for it [23]. Price value has been proved as a determinant for m-banking usage [21]. Different from price value, in this study, cost saving reflects active demand of saving cost for accessing banking service through using m-banking. When users can save some cost for banking services through using mbanking, they would be more willing to use it to conduct various banking practice including UE, UT and UI. Hence, we propose: H2a: Cost saving positively affects UE. H2b: Cost saving positively affects UT. H2c: Cost saving positively affects UI.

3) EFFECTS OF INFORMATION GAINING
Information quality is a strong factor for driving m-banking attitude or usage [9]; providing users with information of high VOLUME 8, 2020 quality (e.g. accurate, relevant, up-to-date and complete information) is a key antecedent of m-banking success. Through providing high quality information, m-banking can satisfy its users' needs and improve their level of usage [1]. Users can use m-banking to gain information of better quality in terms of timeliness, accessibility, personalization, etc. Thus, when a user has a strong demand for gaining information, he/she would be more likely to use m-banking to conduct UE, UT or UI. Hence, we propose: H3a: Information gaining positively affects UE. H3b: Information gaining positively affects UT. H3c: Information gaining positively affects UI.

4) EFFECTS OF WORD-OF-MOUTH
In this study, word-of-mouth reflects that users make use of m-banking because others' positive statement or advice for m-banking. Due to the rise of Web 2.0 technologies, word-of -mouth has become an effective tool for advising product or service in the virtual environment. Ohers' advice or comments are increasingly important for individual's psychological and behavior outcomes [24]. Word-of-mouth can be a kind of social support in the online shopping context [25]. Literature indicates that word-of-mouth can be more powerful than other forms of one-way advertising [26]. Literature has found that word-of-mouth has a positive effect on m-banking attitude and intention [18]. Hence, we propose: H4a: Word-of-mouth positively affects UE. H4b: Word-of-mouth affects UT. H4c: Word-of-mouth affects UI.

5) EFFECTS OF DECISION-MAKING QUALITY
Decision-making quality, which refers to a user's degree of confidence in a purchase decision, is a crucial factor for individual's behavior in the virtual environment [20]. Literature indicates that better purchase decision-making quality can result in higher consumer satisfaction [27]. M-banking is not just a new banking channel that make banking practice more convenient, it can also be a decision support tool for making financial decision. For example, m-banking users can compare different financial products and find out ''the ideal financial product'' that matches their expected return more efficiently and effectively. Hence, when m-banking can help users improve their financial decision-making quality, users would improve UI; however, given the fact that conducting UE and UT do not need much consideration, decision-making quality might not have significant effects on UE and UT. According to the above statement, we propose: H5: Decision-making quality positively affects UI.

B. MEASUREMENT
To measure levels of m-banking usages, we use a set of items that described the frequency of specific m-banking services corresponding to UE, UT or UI [1]. Table 3 shows the wording and descriptive of these items. Measurements for m-banking motivations are adapted from prior literature. Table 4 presents the items for m-banking motivations which are organized in 7-point Likert-style scale (from 1: strongly disagree to 7: strongly agree).

C. PILOT STUDY AND DATA COLLECTION
We translate the items into Chinese and ask two scholars to check on the word choice appropriateness, sentences and content validity. Then, a pilot study with 40 subjects is conducted to assess the measurement properties. Results of Cronbach's coefficient alpha show that all constructs have an acceptable value higher than 0.80. In the formal study, we use an online survey questionnaire method targeted at users of the most popular m-banking in China 2 to collect data. Frist, a screening question (''Are you a user of one or more of the above m-banking?'') and a multiple choice (''Please select the m-banking that you have used'') are used to identify potential respondents. The second section includes the main constructs in the research model. The third section is for asking respondents' demographics including age, gender, education and m-banking experience.
We employ Wjx.cn, a professional online survey company in China, to collect a sample of 600 respondents for formal study. Based on its database, Wjx.cn can randomly distribute questionnaires to the sample group that fits certain conditions (e.g. gender, age, mobile user and other special needs). We use two criteria to identify the invalid questionnaires: 1) time spending on filling in questionnaire < 120 seconds; 2) repeated responses. 92 invalid questionnaires are removed, resulting a valid sample consist of 508 usable responses. Demographic statistics of the sample are shown in Table 5.

1) MEASUREMENT MODEL
First, we test Cronbach's alpha (α) and composite reliability (CR) for each construct. As presented in Table 6, all the α values are larger than 0.8, and all the CR values are larger than 0.9. Thus, the measurement model should have sound internal consistency reliability. Second, we test item loadings and average variance extracted (AVE) to ensure convergent validity. Results show that (Table 6 ) all items load significantly on their corresponding construct with item loadings ranging from 0.860 to 0.953; and AVE values range from 0.753 to 0.880. These results indicate a good convergent  validity for the measurement model. Third, to examine discriminant validity, we calculate the correlation matrix and the square root of AVE of each construct (Table 7 ). Results show that the square root of the AVE of each construct is higher than the correlation between constructs. Therefore, the discriminant validity of measurement model can be confirmed.

2) COMMON METHOD BIAS
Two methods are used to check the extent of CMB in this study. First, we apply the Harman's single-factor method. An exploratory factor analysis with input of all the items is conducted. Results show that the first factor records 22.825% of the total variance, which is lower than the threshold of 50% [29]. Besides, there are five factors with eigenvalues larger than 1, accounting for 75.305% of the total variance. Second, we apply a PLS-SEM based method proposed by [30] to check extent of CMB. Each indicator (item) should be converted to a single-indicator construct, making every construct for m-banking motivation become a second-order construct. Then, we create a common method factor that relates to all the single-indicator constructs. The results indicate that all the single-indicator constructs could be significantly explained by its corresponding constructs but not by  the common method factor. According to the results of these two tests, we conclude that CMB should not be a serious problem in our data.

3) STRUCTURAL MODEL AND HYPOTHESIS TESTING
This article uses the PLS-SEM method to conduct data analysis in study 2. PLS-SEM does not impose sample size restrictions and is distribution-free [31]. PLS-SEM is more appropriate to estimate structural equation model under for VOLUME 8, 2020 small to medium sample sizes than other covariance-based SEM analyses [32]. In addition, PLS-SEM allows for the unrestricted use of single and multiple items constructs [32]. PLS-SEM is chosen in this study because: 1) the sample (n = 508) is a small to medium size; 2) this study has both single and multiple items constructs. Hence, we deem PLS-SEM to be an appropriate technique to test the effects of m-banking motivations on different levels of m-banking usage. FIGURE 2 presents the result of model testing. Convenience can strongly predict UE (β = 0.481, p<0.001) and UT (β = 0.195, p<0.001), supporting H1a and H1b. However, convenience could not significantly predict UI. Thus, H1c is not supported. A plausible reason is: m-banking could help users access their banking services more conveniently, especially those easy day-to-day banking services; both UE and UT are kind of those easy day-to-day banking services, while UI might not be (making an investment requires careful consideration); thus, convenience is an important motivation for UE and UT but not UI.
Cost saving can positively affect UE (β = 0.097, p<0.01) and UT (β = 0.520, p<0.001), supporting H2a and H2b. However, H2c is not supported because cost saving has not significant effects on UI. This result can be attributed to the actual practice of m-banking in China. At present, most of the banks in China have provided free account enquiry service and transfer service via m-banking. Hence, conducting UE and UT can satisfy users' motivation for cost saving. For example, one participant in the qualitative research gives a response ''The main reason is m-banking can provide certain free services, such as accounts transferring and message reminder''.
Information gaining is positively related to UE (β = 0.315, p<0.001) and UI (β = 0.282, p<0.001), supporting H3a and H3c. Hence, users who need to gain information, especially the accurate, relevant, up-to-date and complete banking information, would be more likely to conduct UE and UI via m-banking. However, information gaining does not significantly affect UT, not supporting H3b. This result indicates information gaining is not a crucial motivation for conducting UT in m-banking. The possible reason for this might be that conducting UT in m-banking does not require much information. Word-of-mouth is found to positively affect UT (β = 0.335, p<0.001) and UI (β = 0.192, p<0.001), supporting H4b and H4c. These results suggest that users who take others' recommendation for m-banking might be more likely to conduct UT and UI via m-banking. However, our results indicate that word-of-mouth cannot affect the lowest level of m-banking usage, i.e. UE. The possible reason might be that UE is the basic service of m-banking which is known to all users and does not need other's recommendations.
As expected, decision-making quality can drive UI (β = 0.320, p<0.001). Compared with UE and UT, UI requires users to pay much more attention to consider whether it is right or wrong. That is, decision-making quality is obviously a crucial factor for conducting investmentrelated behavior. When users perceive that m-banking can improve their decision-making quality, they would tend to conduct the investment-related task via m-banking. Hence, the motivation of decision-making quality can increase UI in m-banking.

4) SUMMARIZING AND PRIORITIZING VARIOUS M-BANKING MOTIVATIONS FOR DIFFERENT LEVELS OF M-BANKING USAGE
To address research question 2, Table 8 summarizes the above results. In general, these results suggest that the m-banking motivations identified from study 1 can be used to effectively understand different levels of m-banking; further UE, UT and UI can be driven by different m-banking motivations. In addition, the results could also indicate the priorities of various mbanking motivations for different levels of m-banking usage.
As for explaining UI, the motivations of information gaining (β = 0.282, p<0.001), word-of-mouth (β = 0.192, p<0.001) and decision-making quality (β = 0.320, p<0.001) could take effect. Among these three motivations, decision-making quality (0.320) is the most important motivation, the second is information gaining (0.282), and the last is word-of-mouth (0.192). According to the above results, four m-banking motivations (convenience, cost saving, information gaining and decision-making quality) can significantly drive overall m-banking usage (shown in FIGURE. 3), while five m-banking motivations can significantly drive specific levels of m-banking usage (UE, UT or UI) (shown in FIGURE 2/ Table 8). These findings indicate that treating m-banking usage as homogenous overall usage might obfuscate our understanding on how users make use of m-banking, and disaggregating of different levels of m-banking usage can provide more details and new insights into this issue.

VI. IMPLICATIONS A. THEORETICAL IMPLICATIONS
M-banking usage has gained much attention from research recently. This current study attempts to provide new insights into this issue by investigating different levels of m-banking usage from the perspective of U&G theory via a mixed-method design in China. Our findings can contribute to m-banking literature in two aspects. First, our study can extend m-banking literature by identifying five basic motivations for m-banking usage. As shown in Table 1, existing researches have applied several IS utility theories (e.g. TAM, UTAUT/UTAUT2, D&M IS Success Model, TTF) to examine the factors affecting m-banking usage.
However, the essential motivations for using m-banking has been ignored. Given that individual motivations can provide researchers and practitioners with better understanding on IS usage [12], this study conducts a qualitative study by interviews to identify the basic motivations for m-banking usage, including convenience, information gaining, cost saving, word-of-mouth and decision-making quality. Especially, the motivation of decision-making quality indicates that m-banking is not just a simple mobile channel that makes banking services more convenient, but also a useful tool for improving the effectiveness and efficiency in the financial decision-making process. These m-banking motivations can add to our understanding of why and how individuals make use of m-banking.
Second, this study is one of the first to investigate different levels of m-banking usage from the perspective of U&G theory. Most existing researches on m-banking usage (Table 1) treat m-banking usage as homogenous overall usage, and examine the factors affecting the overall m-banking usage. However, this study argues that different levels of usage might be driven by different factors, especially by different motivations. According to our empirical results of the qualitative study, we confirm that different levels of m-banking usage are driven by different motivations. Specifically, UE can be driven by convenience, cost saving and information gaining; UT can be driven by convenience, cost saving and word-ofmouth; and UI can be driven by information gaining, word-ofmouth and decision-making quality. These findings are in line with literature that claims that investigating specific usage behaviors in IS can enable researchers to gain insights into user motivations of IS use [5].

B. PRACTICAL IMPLICATIONS
Our findings also provide some practical implications for mbanking practitioners. As m-banking being widely accepted, it's of great significance for m-banking service providers to consider how to improve their customers' levels of m-banking usage. In general, our findings highlight the roles of m-banking motivations in improving the levels of m-banking usage. Hence, it is essential for m-banking practitioners to apply the user-centered design rather than the function-centered design to improve users' levels of m-banking usage. In fact, m-banking can serve as a financial toolkit with various services that satisfy different needs. Different strategies should be adopted to improve different levels of m-banking usage.
To improve UE, m-banking service provider should take strategies to enhance the motivations of convenience, information gaining and cost saving for conducting UE. First, convenience is found to be the strongest motivation for UE. Hence, m-banking service provider should focus on how to make UE more convenient, usually in the initial adoption of m-banking. For example, in safety precondition, m-banking service might allow users check their account before login in. Second, information gaining also plays an important role in UE. M-banking providers should focus on improving the ability to provide users with the accurate, relevant, up-to-date and complete account information anytime and anywhere. Last, cost saving can also significantly affect UE. M-banking service should continue to provide free account checking service in m-banking.
To improve UT, our results suggest m-banking providers should try their best to enhance the motivations of cost saving, word-of-mouth and convenience for conducting UT. Cost saving is found be the strongest motivation driving UT. Hence, m-banking providers should provide free transfer service or lower the transfer service fees when their customers conduct UT through m-banking. Besides, results show that UT could be driven by word-of-mouth. This finding indicates that UT might be one of the main advantages of m-banking over other banking channel. M-banking providers should take some measures to encourage their old users to promote this advantage. In addition, m-banking might provide diversified payment method (such as fingerprint payment, face payment and voice payment) and automatically record and match the UT information to improve the convenience of UT, as convenience can increase UT.
Most importantly, in terms of UI, the highest level of mbanking usage in this study, m-banking service providers should pay attention to users' motivations of decision-making quality, information gaining and word-of-mouth. To enhance decision-making quality, useful investment analysis services should be provided within m-banking, which could help users select the best financial products effectively and efficiently. In addition, to enhance the motivation of information gaining, m-banking provider should provide users with useful financial information to help customer conduct UI. Besides, word-of-mouth is also a motivation for UI, which means that m-banking providers should also encourage their old users to share the advantage of conducting UI in m-banking.

VII. CONCLUSION
In conclusion, this article focuses on the levels of m-banking usage from the perspective of U&G theory. First, this study provides three levels of m-banking usage, namely UE, UT and UI. Then, this study applies a mixed-method design to investigate the effects of m-banking motivations on these three levels of m-banking usage. Based on the qualitative study (Study 1), five motivations for m-banking usage are identified, namely convenience, cost saving, information gaining, word-of-mouth and decision-making quality. Drawing from the U&G theory, results of the quantitative study (Study 2) indicate that different levels of m-banking usage could be driven by different m-banking motivations. We believe this study can provide some contributions to m-banking literature and some practical implications to the m-banking service providers.

APPENDIX A
See Table 9.