International Journal of Bank Marketing Emerald Article: Mobile banking: proposition of an integrated adoption intention framework Júlio Püschel, José Afonso Mazzon, José Mauro C. Hernandez Article information: To cite this document: Júlio Püschel, José Afonso Mazzon, José Mauro C. Hernandez, (2010),"Mobile banking: proposition of an integrated adoption intention framework", International Journal of Bank Marketing, Vol. 28 Iss: 5 pp. 389-409 Permanent link to this document: http://dx.doi.org/10.1108/02652321011064908 Downloaded on: 31-05-2012 References: This document contains references to 50 other documents To copy this document: permissions@emeraldinsight.com This document has been downloaded 2130 times since 2010. * Users who downloaded this Article also downloaded: * Nicole Koenig-Lewis, Adrian Palmer, Alexander Moll, (2010),"Predicting young consumers' take up of mobile banking services", International Journal of Bank Marketing, Vol. 28 Iss: 5 pp. 410-432 http://dx.doi.org/10.1108/02652321011064917 Hernan E. Riquelme, Rosa E. Rios, (2010),"The moderating effect of gender in the adoption of mobile banking", International Journal of Bank Marketing, Vol. 28 Iss: 5 pp. 328-341 http://dx.doi.org/10.1108/02652321011064872 Lisa Wessels, Judy Drennan, (2010),"An investigation of consumer acceptance of M-banking", International Journal of Bank Marketing, Vol. 28 Iss: 7 pp. 547-568 http://dx.doi.org/10.1108/02652321011085194 Access to this document was granted through an Emerald subscription provided by UNIVERSITI SAINS MALAYSIA For Authors: If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service. Information about how to choose which publication to write for and submission guidelines are available for all. Additional help for authors is available for Emerald subscribers. Please visit www.emeraldinsight.com/authors for more information. About Emerald www.emeraldinsight.com With over forty years' experience, Emerald Group Publishing is a leading independent publisher of global research with impact in business, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, as well as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation. *Related content and download information correct at time of download.
The current issue and full text archive of this journal is available at www.emeraldinsight.com/0265-2323.htm Mobile banking: proposition of an integrated adoption intention framework Júlio Püschel and José Afonso Mazzon Faculdade de Economia, Administração e Contabilidade (FEA/USP), Universidade de São Paulo, São Paulo, Brazil, and José Mauro C. Hernandez Centro Universitário da FEI, São Paulo, Brazil Mobile banking 389 Abstract Purpose This paper s objective is to propose an integrated framework to investigate the adoption intention of mobile banking technology and to test it in the Brazilian context. Design/methodology/approach A total of 666 respondents from the most economically developed cities in Brazil were surveyed. The sample comprised 333 mobile banking users and 333 mobile banking non-users. Partial least squares was used to analyze the proposed framework s construct relations. Findings The framework offers an integrated view, taking into account more predictors than other studies on the adoption of innovations. For non-users, the framework was able to explain approximately 69 percent of the dependent variable (intention to adopt mobile banking) variation, which is a figure higher than those obtained in previous studies. However, for current users of mobile banking, only 27 percent of the dependent variable variation was explained by the framework. It was also observed that the predictors influence over the criterion variable was different for each group of mobile banking users and non-users. Originality/value The findings suggest that the proposed integrated framework offers a deeper understanding of the variables that influence the adoption of mobile banking. Keywords Mobile communication systems, Banking, Innovation, Consumer behaviour, Least square approximation, Brazil Paper type Research paper Introduction Internet evolution is changing the way people communicate and interact with others within their social circle. The growing demand for always-on internet connectivity is shifting usage from desktops to laptops and mobile devices. Telecommunications are also migrating not only from fixed to mobile but also from voice to data, allowing people to be connected anywhere, anytime. In order to meet customer expectations, financial services companies are pursuing alternative channels in order to increase customer convenience, reduce costs and maintain profitability. Phone banking and automated teller machines (ATM) are already widely used by financial institutions in many countries and are becoming everyday more sophisticated (White, 1998). In Brazil, where this study was carried out, there are more than 170,000 ATMs, a number significantly higher than the 19,000 bank branches. Internet banking is also a channel that has gained the attention of most Brazilian financial institutions, since there are more than 32 million bank clients using internet banking (FEBRABAN, 2008). International Journal of Bank Marketing Vol. 28 No. 5, 2010 pp. 389-409 q Emerald Group Publishing Limited 0265-2323 DOI 10.1108/02652321011064908
IJBM 28,5 390 As a result of initiatives to increase customers banking convenience, financial services companies are launching banking services also through a mobile network. This service, called mobile banking, has been defined as a way of executing financial services through the use of mobile communications technology (Pousttchi and Schurig, 2004). Mobile banking has high potential, since it follows on the success of internet banking (Brown et al., 2003). Therefore, mobile banking is not only a natural evolution of internet banking, but it is also a better digital alternative to other traditional bank channels such as ATMs, internet banking and physical branches. Estimates of the number of mobile banking users confirm this prognosis: there are approximately six million people performing financial transactions through mobile banking in Western Europe (Riivari, 2005) and something like 700,000 mobile banking users in Latin America, a number which should reach 4 million in 2012 (Püschel, 2008). In addition, mobile phones are increasing at an annual rate of 14 percent, and now total 3.8 billion around the world (Yankee Group, 2008). Consequently, the interest of financial institutions to offer applications and services via the mobile channel is increasing at the same speed as cell phone penetration. Recent literature (Scornavacca and Hoehle, 2007; Herzberg, 2007; Mallat et al., 2004; Riivari, 2005; Suoranta and Mattila, 2004) has shown an interest in investigating the adoption of this new banking channel. However, most mobile banking studies have applied a wide variety of methods and frameworks, making it difficult to compare the results of different studies and evolve an adoption theory in this field. The objective of the present study is to propose an integrated mobile banking adoption framework by focusing on the factors that influence consumer decisions as to whether to adopt mobile banking. The proposed integrated framework is based on consolidated social psychology theory and well-known innovation diffusion and technology adoption frameworks. The main advantage of this framework is the consolidation of a greater number of predictor variables that are able to provide a fuller explanation of the consumer s adoption of mobile banking. In order to meet the objective, the paper was organized in eight sections. In the first section, we describe the literature review, a brief overview of the adoption innovation models and the proposed integrated adoption intention framework. Then we provide a description of the constructs included in the proposed framework. In the following sections we describe the proposed method used to test the framework and the analysis results. Finally, the study concludes by putting forward some conclusions and suggestions for future endeavors in this area. Literature review There is a growing interest in studying the adoption of mobile banking (Brown et al., 2003; Lee et al., 2003; Suoranta and Mattila, 2004; Laforet and Li, 2005; Luarn and Lin, 2005; Pedersen, 2005; Sulaiman et al., 2007; Laukkanen et al., 2008), but literature on the subject is still in its infancy. Previous studies suggest the use of some of the theories from information systems research and innovation diffusion, but there is still no standard of how to apply these theories in the mobile banking context. For example, Sulaiman et al. (2007) and Brown et al. (2003) applied Roger s innovations diffusion theory to investigate mobile banking adoption, while Luarn and Lin (2005) opted for Davis (1989) Technology Acceptance Model (TAM) and Pedersen (2005) applied
Taylor and Todd s (1995) Decomposed Theory of Planned Behavior (DTPB). Other authors used no innovation adoption models (Lee et al., 2003; Laforet and Li, 2005; Laukkanen et al., 2008). The relatively small number of studies and the variety of models used in previous studies justify studying mobile banking adoption in more depth and the endeavor of proposing an integrated model that provides a broader explanation of the phenomenon. By on the other hand, when analyzing innovation adoption models (Rogers, 1962) and, more specifically, the adoption of technology innovation models (Davis, 1989; Moore and Benbasat, 1991; Taylor and Todd, 1995), it is clear that despite their similarities, these models consider distinct groups of predictive variables. Taken individually each model has a limited predictive power over the adoption of technology innovations. The idea behind this study is that integrating these different models into one single framework not only offers greater predictability when compared with each model taken individually, but also makes it possible to identify the cross-correlation of the predictive variables. The proposed theoretical framework departs from the DTPB proposed by Taylor and Todd (1995) and includes elements of the technology acceptance model (Davis, 1989) and the innovation diffusion theory (Rogers, 1962). In the following section, we present the adoption of the innovations models that were used as support for the framework proposed in this study. Mobile banking 391 A brief overview of the adoption of innovations models The Innovations Diffusion Theory (IDT), proposed by Rogers (1983), is the best-known model of innovation adoption. According to this model, the adoption rate of a new technology depends on five characteristics of the innovation, namely relative advantage, compatibility, complexity, observability, and trialability. Except for the construct complexity, which has a negative relationship with the adoption rate, the other constructs positively affect adoption intention. When developing an instrument to measure initial adoption and possible diffusion of information technology innovations within organizations, Moore and Benbasat (1991) added two other constructs to Rogers s (1983) model: image and voluntariness of use. Furthermore, Moore and Benbasat (1991) broke down construct observability into two: results demonstrability and visibility. This instrument tries to capture perceptions of the innovation characteristics, since it is more likely that the behavior of those who adopt it in the future will be more dependent on their perceptions of the innovation. While Rogers (1983) model is based on the characteristics of the innovation, many innovation adoption models and theories are based on two well-known social psychology theories: the Theory of Reasoned Action (TRA) (Ajzen and Fishbein, 1970) and its extension, the Theory of Planned Behavior (TPB) (Ajzen, 1985). In TPB, behavior is determined by both behavioral intention and perceived behavioral control, while behavioral intention is influenced by the attitude towards behavior, the subjective norm, and the perceived behavioral control. The TAM was proposed by Davis (1989) to predict the adoption of new information technologies within organizations and is based on TRA. In TAM, behavioral intention is determined by both the attitude towards the use of the system and its perceived usefulness. In turn, attitude is influenced by both the perceived usefulness of the system and by the perceived ease of use. Later, Venkatesh and Davis (2000) included the construct subjective norm in the original model and called it TAM2.
IJBM 28,5 392 In the context of information technology adoption, Taylor and Todd (1995) advanced the DTPB. They departed from the TPB and decomposed subjective norm, perceived behavioral control and attitude constructs into new sub-constructs. Specifically, subjective norm was split into peer influence and superior influence, perceived behavioral control was decomposed into self-efficacy, resources facilitating conditions, and technology facilitating conditions, while the attitude construct was split into three dimensions: two of them were based on TAM variables, perceived usefulness and ease of use, and the third one, compatibility, was based on Roger s (1962) innovation diffusion theory. According to Taylor and Todd (1995), the main advantages of the DTPB are to provide a better understanding of the relationship between attitudes and intentions and to be more managerially-oriented, since it reveals specifically which predictors are more relevant in explaining adoption behavior. To date, DTPB is the most complete model of innovation adoption yet developed and widely used in different fields of knowledge. In the next section we detail the proposed integrated model of mobile banking intention adoption. The proposal of an integrated model of mobile banking intention adoption We propose to integrate elements of the best known models of innovation adoption to put forward a framework for predicting mobile banking intention adoption. We adjust DTPB to the mobile banking context and add elements from other theories as detailed below:. Attitude towards intention to adopt are decomposed into seven sub-constructs - compatibility and perceived ease of use were retained from DTPB, and we included five others, namely, relative advantage and trialability from IDT, and image, visibility, and results demonstrability from Moore and Benbasat (1991).. Perceived behavioral control was decomposed into three sub-constructs, as in DTPB.. Subjective norm was not decomposed as in the DTPB, since the adoption of mobile banking is a personal decision rather than a decision taken in the organizational context, as proposed by DTPB.. We examined both intention to adopt and actual adoption as main dependent variables of the model. To sum up, the proposed integrated framework is shown in Figure 1 and considers eleven independent variables: relative advantage (RA), compatibility (CO), image (IM), results demonstrability (RD), trialability (TR), visibility (VI), perceived ease of use (PEU), self-efficacy (SE), resource facilitating condition (RFC), technology facilitating condition (TFC) and subjective norm (SN). The dependent variables of the framework are attitudes (AT), perceived behavioral control (PBC) and intention (IN)/usage (US) of mobile banking. The relationships between each variable are further discussed below. Attitude is defined by Allport (1935) as a mental and neural state of readiness, organized through experience, exerting a directive or dynamic influence upon the individual s response to all objects and the situations with which they are related. Ajzen and Fishbein (1980) state that attitudes are the beliefs an individual has related to the results that the adoption of a specific behavior will offer and his evaluation of the
Mobile banking 393 Figure 1. Proposed integrated framework possible outcomes. In TRA, the more positive the attitude towards the behavior; the greater the intention to perform a specific behavior. Therefore, we hypothesize that the better the attitudes towards the use of mobile banking, the greater the intention to adopt or continue using mobile banking. Relative advantage is defined as the degree to which an innovation is perceived as a better alternative to currently available products or services (Rogers, 1962) and can be related to diverse economic, social, convenience, and satisfaction dimensions. The inclusion of relative advantage in the model precluded the inclusion of TAM s perceived usefulness, since they are similar in content, but relative advantage is more comprehensive, encompassing more diverse dimensions than perceived usefulness. In the proposed framework, the more one perceives the relative advantages of mobile banking, the better the attitudes towards it. Rogers (1962) saw image as a dimension of the relative advantage construct, but Tornatzky and Klein (1982) argued that image has an important effect on attitudes and needed to be considered as a separate variable from relative advantage. Moore and
IJBM 28,5 394 Benbasat (1991) defined image as the degree to which use of an innovation is perceived to enhance one s image or status in one s social system and found that, in fact, it has an impact on the adoption of innovations. Based on previous results, we state that the more one perceive mobile banking as being compatible with one s own image, the better one s attitudes towards mobile banking will be. Rogers and Shoemaker (1971) defined trialability as the degree to which an innovation can be tried on a limited basis. Consequently, new ideas that can be tested before their full implementation are normally adopted faster than those that cannot be tried (Rogers, 1962). The proposed framework considers trialability as the perception of individuals about how much banks offer chances for them to try mobile banking services and, the clearer this perception, the better their attitudes towards mobile banking. Visibility and results demonstrability have their origin in Roger s (1962) observability construct. Visibility refers to the extent that an innovation can be observed before it is adopted (Moore and Benbasat, 1991), while results demonstrability refers to the degree to which the benefits of using an innovation are clear to potential adopters (Moore and Benbasat, 1991). Therefore, we hypothesize that the more visible mobile banking is and the more individuals are able to perceive its benefits, the better the attitudes towards mobile banking. Rogers (1962) defined compatibility as the degree to which an innovation is perceived as being consistent with the individual s values, past experiences, and needs. Therefore, we hypothesize that the more mobile banking is perceived as being compatible with such values, experiences and needs, the better the attitudes towards it. The concept of perceived ease of use was first proposed by Davis (1989) and defined as the degree to which a person believes that using a particular system would be effortless. This construct was included in DTPB and retained in our proposed framework. Hypothetically, the greater the perception that mobile banking is easy to use, the better the attitudes towards it. Subjective norm refers to the perceived social pressure about whether to adopt a specific behavior (Ajzen, 1985). Social pressure can be exerted by family, friends or individuals that belong to the same social groups as the potential adopter. Although subjective norm was split into two constructs by Taylor and Todd (1995), in this study we kept it as a single construct, because mobile banking adoption is basically a personal decision rather than an organization-dependent decision. In the mobile services context, Pedersen (2005) found that subjective norm is an important predictor of intention to use mobile commerce. We hypothesize that the greater the perception of social pressures towards the use of mobile banking, the greater the intention to adopt it, or continue using it. Perceived behavioral control is defined as the resources and opportunities available to an individual that offer the conditions necessary for adopting a certain behavior Ajzen (1991). Taylor and Todd (1995) decomposed the construct into self-efficacy, resource facilitating conditions and technology facilitating conditions. In the proposed framework, perceived behavioral control is based on the potential adopter s perception about whether he is capable of using mobile banking and possesses the required knowledge and resources to adopt the service. The hypothesis is that once an individual perceives that he is able to use mobile banking and the technological and other resources that are available to him, the more likely it is he will use, or continue using mobile banking.
Behavioral intention came from Ajzen and Fishbein s (1969) theory of reasoned action and serves as an intervening variable between behavior and attitude. They argue that behavioral intention is responsible for a large part of behavior variance, making it possible to predict specific behaviors based on intention to perform the behavior. In this study we defined the behavioral intention as the intention to adopt or continue using mobile banking. This definition required us to collect data from both users and non-users of mobile banking. For mobile banking non-users, behavioral intention was measured as the intention to adopt the service and recommend it to other people while for mobile banking users, behavioral intention projects the intention to maintain current use and usage frequency. Mobile banking 395 Method and procedures Implementation of the method proposed in the previous section for determining the factors that influence mobile banking adoption is described below. The research took place in Brazil with both mobile banking user and non-user databases. Both the survey and data analysis methods will be described in the following sections. Survey methodology and questionnaire The survey was based on a structured online questionnaire with 666 respondents, collected in the first semester of 2009. Although there were more respondents than the current sample indicates, all responses were checked for consistency and a few were dropped during this process. Therefore, it was decided to use the same sample size for both groups (333 users and 333 non-users). The respondents involved in the survey have at least one banking account and were also sampled by convenience, in order to represent the major economic cities in Brazil. Two types of questionnaire were applied in order to differentiate mobile banking users from non-users. The major difference was the intention/behavior scale, which was applied differently, depending on the group; an intention scale was used for non-users and a behavior scale for users. As suggested by Moore and Benbasat (1991), the items of each construct were formulated in terms of the perception of performing/adopting mobile banking services. The questionnaire dealt with similar topics to allow for their investigation within different populations. In this case, all constructs were expressed in a different verb tense, according to the mobile banking users status (for users, verbs were used in the present or past tense, while for non-users verbs were in a conditional tense). All scales used to measure the framework constructs were based on Moore and Benbasat (1991) for compatibility, relative advantage, visibility, results demonstrability, image and trialability; Davis (1989) for perceived ease of use and Taylor and Todd (1995) for attitude, subjective norm, perceived behavioral control, self-efficacy, resource facilitating condition, technology facilitating condition and intention. Adjustments were made in some of the phrases in order to adapt their meaning to fit the mobile banking context. Furthermore, all items of the constructs were measured on a six-point Likert-type scale, ranging from 1 totally disagree to 6 totally agree. The scales were applied in a two-step process, in which the respondents first answer if they agree or disagree with the scale item and then state their level of agreement or disagreement, ranging from somewhat to totally agree/disagree (Mazzon, 1981).
IJBM 28,5 396 Data analysis The partial least squares (PLS) method was used to analyze the study results in order to consider the influence of all constructs on the framework simultaneously, including the second order construct (intention/behavior). Fornell and Bookstein (1982) argue that PLS should divide the model parameters into two subgroups, applying several regressions to estimate these parameters. Moreover, Gudergan et al. (2008) describe PLS as being a sustainable method for estimating cause and effect relationships in complex business research. Hwang et al. (n.d.) state that the PLS method is particularly applicable to contexts in which there are no sufficiently robust and well-structured theories. The PLS approach is able to help obtain values for latent variables for predictive purposes. PLS does not use a model for explaining the covariances of all indicators; it minimizes the variance of all dependent variables, based on the parameter estimates obtained, which in turn are based on the ability to minimize the residual variance of the dependent variables (Chin, 1998a). In addition, Chin et al. (1996) argue that PLS estimates, reflecting the latent variables by their indicators, result in a more accurate analysis of constructs and their relationships. The authors also explain that PLS is widely used because of its ability to model latent variables under non-normality conditions, with small and medium-sized samples. The authors also argue that PLS, unlike multiple regression, does not assume equal weights for all items in a scale, but allows each indicator to vary according to its contribution weigh to the latent variable. Chin (1998b) states that SEM techniques are among the most well-known covariance-based methods and that many social science researchers consider these methods as being tautologically synonymous to PLS. However, the author states that PLS is an alternative technique for researchers interested in doing SEM-based analysis. PLS can be argued to be more suitable, depending on the researcher s objective and epistemic view of data to theory, the properties of the data in-hand and measurement development. Furthermore, Stan and Saporta (1995) describe that in the PLS approach, there are fewer probabilistic hypotheses, data are modeled by a succession of simple or multiple regressions and there is no identification problem. In addition, the authors state that in the SEM (LISREL) approach, estimation is done by maximum likelihood, based on the hypotheses of multi-normality, and allows the variance matrix to be modeled. PLS can be a powerful method of analysis because of its minimal demands on measurement scales, sample size and residual distributions (Chin, 1998b). In addition, Stan and Sapporta (1995) state that, on the one hand, PLS is prediction-oriented and optimal for prediction accuracy, supports highly-complex models and is more flexible, while on the other, SEM is oriented towards parameter estimation and optimal for parameter accuracy, supports small, moderately complex models and relies on key assumptions. Moreover, Chin et al. (1996) state that LISREL is superior to PLS because of its ability to estimate the underlying population parameters; however, this becomes less of a concern if the objective is to account for multi-variance in a predictive sense. In this study, the choice to use the PLS method is based on the fact that the proposed theoretical framework includes a large complexity in its variables and relations and because the model aims to predict future behavior rather than estimate population parameters. Finally, the software used was the SmartPLS 2.0 (Ringle et al., 2005).
Results Sample characteristics Mobile banking users are predominantly young males who have had a bank account for less time than non-mobile banking users, as shown in Table I. Regarding mobile banking usage, the results show that the service is mostly restricted to account balance checking and bill payments, as shown in Table II. Construct validity, dimensionality and reliability In order to access the construct s validity, dimensionality and reliability, users and non-users of mobile banking were analyzed as two separate groups. For each construct the average variance extracted (AVE), its square root, composite reliability, R 2, Cronbach s alpha and communality were calculated. Mobile banking 397 Mobile banking non-users Construct validity was obtained by first accessing Cronbach s alpha for each construct. The results are shown in Table III. With the exception of perceived behavior control, all other constructs obtained a higher alpha than the recommended 0.7 (Hair et al., 2005). Next, all constructs were processed in the model via PLS assessment in order to obtain each construct s AVE, composite reliability and communality. All constructs obtained more than the minimum required for each parameter (Chin 1998a, b; Stan and Saporta, 2005). Then, based on Geffen and Staub (2005), each construct had its AVE square root extracted in order to assess construct dimensionality. The results were used as a reference when all constructs were correlated and the weight of each correlation within the two constructs needs to be lower than the square root of the AVE as shown in Table IV. Variable Category Users (%) Non-users (%) Age # 30 56 44 31-45 38 34 $ 46 6 22 Gender Male 78 56 Female 22 44 Time as banking customer (years) # 10 49 38 11-25 37 39 $ 26 14 23 Table I. Sample demographic characteristics Types of service % Account balance checking 52 Bill payments 24 Buying mobile pre-paid credits 16 Transferring amounts 10 Credit card balance checking/payments 7 General banking requirements 5 Table II. Mobile banking service usage
IJBM 28,5 Constructs AVE AVE square root Composite Cronbach s reliability R 2 alpha Communality 398 Table III. Construct reliability and validity for non-users of mobile banking Attitude 0.723 0.850 0.912 0.679 0.871 0.723 Self-efficacy 0.645 0.803 0.845 0.724 0.645 Compatibility 0.765 0.874 0.928 0.897 0.765 Perceived behavioral control 0.600 0.775 0.818 0.628 0.670 0.600 Results demonstrability 0.669 0.818 0.890 0.836 0.669 Technology facilitation condition 0.546 0.739 0.826 0.715 0.546 Resource facilitation condition 0.577 0.760 0.843 0.749 0.577 Perceived ease of use 0.588 0.767 0.850 0.769 0.588 Image 0.621 0.788 0.866 0.796 0.621 Intention 0.891 0.944 0.970 0.686 0.959 0.891 Subjective norm 0.774 0.880 0.911 0.856 0.774 Testability 0.658 0.811 0.885 0.828 0.658 Relative advantage 0.636 0.797 0.896 0.854 0.636 Visibility 0.658 0.811 0.885 0.826 0.658 Only two correlations obtained a value greater than the AVE s square root (image and subjective norm; self-efficacy and perceived behavioral control). In order to refine the model, all correlations greater than the AVE s square root were assessed in order to correlate the indicators of these constructs. After analyzing the correlation of the construct indicators, those indicators with a high correlation were eliminated (one indicator from perceived behavioral control, one from self-efficacy, one from image and one from subjective norm) and a new AVE was calculated. Finally, all construct correlations were shown to be lower than the new AVE s square root after refining the model and all parameters (AVE, composite reliability, Cronbach s alpha and communality) obtained satisfactory indices. The results are shown in Table IV. Mobile banking users Mobile banking user results also obtained higher alphas than the recommended 0.7 (Hair et al., 2005) with the exception of perceived behavioral control. In addition, all constructs obtained more than the minimum required for each parameter: AVE, composite reliability and communality (Chin 1998a, b; Stan and Saporta, 2005), as described in Table V. The same process applied to mobile banking non-users was used for users, based on the approach of Geffen and Staub (2005). Three correlations resulted in weights that were higher than the AVE square root (attitude and perceived ease of use; perceived behavioral control and self-efficacy; self-efficacy and perceived ease of use; subjective norm and image). Therefore, after analysis of each of the construct indicators, those with high correlations were then eliminated (one from attitude, one from perceived ease of use, two from self-efficacy and one from subjective norm) and a new AVE was calculated, resulting in no correlation being higher than each AVE s square root, as shown in Table VI. Path analysis model testing The 13 paths were analyzed in order to access the effect size ( f 2 ) to distinguish which of the paths contribute towards explaining the dependent variable to which they are
Constructs AT SE CO PBC RD TFC RFC PEU IM IN NS TR RA VI Attitude 0.85 Self-efficacy 0.67 0.80 Compatibility 0.74 0.66 0.87 Perceived behavioral control 0.48 0.68 0.54 0.78 Results demonstrability 0.66 0.62 0.73 0.61 0.82 Technology facilitation condition 0.31 0.50 0.41 0.69 0.43 0.74 Resource facilitation condition 0.40 0.53 0.42 0.53 0.48 0.65 0.76 Perceived ease of use 0.44 0.73 0.49 0.73 0.60 0.50 0.45 0.77 Image 0.41 0.26 0.51 0.32 0.42 0.24 0.25 0.16 0.79 Intention 0.80 0.63 0.81 0.46 0.68 0.52 0.36 0.40 0.47 0.94 Subjective norm 0.41 0.28 0.48 0.31 0.41 0.28 0.23 0.14 0.76 0.50 0.88 Testability 0.55 0.54 0.63 0.41 0.60 0.34 0.49 0.41 0.37 0.59 0.26 0.81 Relative advantage 0.75 0.54 0.66 0.38 0.63 0.31 0.36 0.38 0.39 0.70 0.39 0.55 0.80 Visibility 0.63 0.51 0.67 0.40 0.63 0.28 0.39 0.33 0.53 0.65 0.45 0.71 0.63 0.81 Mobile banking 399 Table IV. Variable correlation matrix based on the AVE square root for non-users of mobile banking
IJBM 28,5 Constructs AVE AVE square root Composite Cronbach s reliability R 2 alpha Communality 400 Table V. Construct reliability and validity for mobile banking users Attitude 0.699 0.836 0.874 0.756 0.782 0.699 Self-efficacy 0.724 0.851 0.840 0.623 0.724 Compatibility 0.617 0.785 0.865 0.792 0.617 Perceived behavioral control 0.602 0.776 0.818 0.771 0.665 0.602 Results demonstrability 0.648 0.805 0.880 0.819 0.648 Technology facilitation condition 0.558 0.747 0.831 0.723 0.558 Resource facilitation condition 0.589 0.768 0.846 0.753 0.589 Perceived ease of use 0.591 0.769 0.847 0.758 0.591 Image 0.629 0.793 0.892 0.846 0.629 Use 0.760 0.872 0.927 0.279 0.896 0.760 Subjective norm 0.657 0.810 0.852 0.747 0.657 Testability 0.589 0.767 0.846 0.755 0.589 Relative advantage 0.582 0.763 0.872 0.816 0.582 Visibility 0.562 0.750 0.834 0.734 0.562 attached. Chin (1998b) states that the R 2 for each latent variable (LV) can be a starting point when analyzing PLS for the structured model, since interpretation of the PLS is similar to that of a traditional regression. The author also states that the change in the R 2 can be explored to see whether the impact of a particular independent LV on a dependent LV is substantial. Following Chin s (1998b) recommendation, effect size can be calculated as: f 2 ¼ R 2 included 2 R 2 excluded 1 2 R 2 included where R 2 included and R 2 excluded are the R 2 provided on the dependent LV, when the predictor LV is used or omitted from the structural equation, respectively. The f 2 of 0.02, 0.15 and 0.35 can be interpreted as a predictor LV having a small, medium, or large effect at the structural level. Then the Q 2 values were accessed in order to identify cross-validation and consequently, predictive relevance. This procedure was based on the predictive sample reuse technique from Stone (1974) and Geisser (1975). Chin (1998b, p. 116) states that the PLS adaptation of this approach follows a blindfolding procedure that omits a part of the data for a particular block of indicators during the parameter estimation and attempts to estimate the omitted part using the estimated parameter. The predictive measure for the block becomes: P Q 2 ¼ P DE D where D is the omission distance, E is the prediction error, calculated when the omitted data points are subsequently predicted, and O is the square errors using the mean for prediction. Therefore, without any loss of freedom, Q 2 represents a measure of how well-observed values are reconstructed by the model and its parameter estimates. D o D
Constructs AT SE CO PBC RD TFC RFC PEU IM IN NS TR RA VI Attitude 0.84 Self-efficacy 0.60 0.85 Compatibility 0.71 0.31 0.79 Perceived behavioral control 0.55 0.74 0.34 0.78 Results demonstrability 0.56 0.47 0.63 0.50 0.81 Technology facilitation condition 0.47 0.44 0.36 0.73 0.39 0.75 Resource facilitation condition 0.30 0.45 0.18 0.63 0.30 0.58 0.77 Perceived ease of use 0.70 0.70 0.53 0.68 0.52 0.55 0.39 0.77 Image 0.26 0.21 0.30 0.26 0.51 0.17 0.36 0.17 0.79 Intention 0.41 0.30 0.31 0.41 0.51 0.38 0.51 0.26 0.49 0.87 Subjective norm 0.22 0.13 0.19 0.09 0.50 0.04 0.13 0.10 0.76 0.32 0.81 Testability 0.57 0.38 0.71 0.38 0.60 0.34 0.20 0.58 0.41 0.20 0.23 0.77 Relative advantage 0.69 0.31 0.64 0.35 0.30 0.33 0.23 0.48 0.36 0.28 0.23 0.37 0.76 Visibility 0.61 0.42 0.59 0.37 0.66 0.27 0.26 0.40 0.62 0.40 0.45 0.73 0.46 0.75 Mobile banking 401 Table VI. Variable correlation matrix based on the AVE square root for mobile banking users
IJBM 28,5 402 Q 2. 0 implies the model has predictive relevance and Q 2, 0 represents a lack of predictive relevance. The results of Q 2, f 2 and path coefficients will be described in the following section, divided into mobile banking user and non-user groups. Mobile banking non-user group From the 13 paths that form an integral part of the conceptual framework, six had an effect over the dependent LV. Of those, only Attitude! Intention had a large effect (f 2. 0:35); Technology facilitation condition! Perceived behavioral control; Self-efficacy! Perceived behavioral control and Relative advantage! Attitude had a medium effect; Compatibility! Attitude and Subjective Norm! Intention had a small effect. Although all paths were relevant based on a Q 2. 0, paths with f 2, 0:02 were considered not significant as their effect size in the proposed model is almost null. Attitude resulted in the largest influence on LV intention, with a path coefficient of 0.657. Subjective norm had the second largest coefficient of 0.141 (small effect), which is coherent with previous behavioral prediction theories (Ajzen and Fishbein, 1969; Ajzen, 1985; Davis, 1989; Taylor and Todd, 1995). Both LVs were responsible for explaining almost 68 percent of intention variance, which is a high rate compared to 40 percent from previous studies on technology adoption (Venkatesh et al., 2003) Technology facilitation condition and self-efficacy had significant path coefficients over Perceived behavioral control, as argued by Taylor and Todd (1995); however, Perceived behavioral control had an almost null effect size on intention, which may be a consequence of the adoption of mobile banking being beyond individual control, leading to irrelevant perceived behavior control In previous technology adoption studies, perceived ease of use was the second most relevant factor after attitude. In this study, compatibility with life-style had the second highest coefficient that preceded attitude, representing a possible peculiarity of mobile phone technology. As mobile services are normally personal, their compatibility with life-style becomes an important factor. However, other subjective factors such as results demonstrability and image had almost no influence over attitude. Again, as mobile services are normally very personal, the results of using the services are important to the user himself and how others perceive his using the service is not relevant. Table VII shows that six of the 13 hypotheses from the proposed framework cannot be rejected (see Figure 2). Mobile banking users The framework applied to banking users resulted in only three paths with a medium and large effect size. Perceived ease of use (Path coefficient ¼ 0:505) and relative advantage (PC ¼ 0:458) were the two constructs with the highest path coefficients of all 13 paths. However, Perceived ease of use resulted in a medium effect size (f 2 ¼ 0:270), while relative advantage had a small effect size (f 2 ¼ 0:135). The core framework constructs attitudes (f 2 ¼ 0:037), subjective norm (f 2 ¼ 0:080) and Perceived behavioral control (f 2 ¼ 0:074) obtained a small effect size in the intention construct, which indicates that the proposed framework might be useful only for predicting future behaviors but not the possibility of continuing to adopt the same behavior, in this case, continuing to use mobile banking. Technology facilitating condition and self-efficacy had a large effect size and predictive relevance on the Perceived behavioral control construct, with an outcome
Constructs Path coefficients LV index values f 2 Q 2 Mobile banking Attitude! Intention 0.657 4.141 1.041 0.488 Technology facilitating condition! Perceived behavioral control 0.574 3.719 0.301 0.546 Self-efficacy! Perceived behavioral control 0.503 4.400 0.341 0.645 Relative advantage! Attitude 0.459 3.963 0.263 0.636 Compatibility! Attitude 0.342 3.940 0.119 0.765 Subjective norm! Intention 0.141 2.367 0.092 0.774 Visibility! Attitude 0.106 4.090 0.013 0.658 Results demonstrability! Attitude 0.076 4.102 0.006 0.669 Perceived behavioral control! Intention 0.051 4.070 0.016 0.376 Perceived ease of use! Attitude 0.048 4.635 0.006 0.588 Image! Attitude 20.004 2.911 0.003 0.621 Trialability! Attitude 20.026 4.661 0.003 0.658 Resource facilitating condition! Perceived behavioral control 20.033 4.098 0.003 0.573 403 Table VII. Path analysis mobile banking non-users Figure 2. Mobile banking integrated adoption framework - non-users
IJBM 28,5 404 path coefficient of 0.416 (Technology facilitating control) and 0.373 (Self-efficacy). However, their influences are minimized by the small effect size from Perceived behavioral control on the intention to continue using mobile banking construct. Trialability and images resulted in negative path coefficients, which is different to the expected positive outcomes, based on Rogers (1962), Tornatzky and Klein (1982) and Moore and Benbasat (1991). Although the small effect size indicates that the influence of these constructs on the LV attitude is low, other factors could be indicators of the construct s negative coefficient. Image s low influence might be the result of the respondent not wanting to admit during the interview that image is an important factor. Additionally, trialability s low influence might be a consequence of that fact that it is irrelevant to try the service if the adopter is evaluating whether to continue using it. Finally, six of the 13 hypotheses from the proposed framework applied to mobile banking users cannot be rejected. However, only three can be considered to have a medium and large effect size on the dependent LV, as shown in Table VIII (see also Figure 3). Discussion The conceptual framework proposed an evolution of the DTPB (Taylor and Todd (1995), adding additional construct antecedents to attitude, based on Rogers (1962) and Moore and Benbasat (1991); literature shows that these scales are normally used individually. Hernandez and Mazzon (2007) have jointly used these scales to evaluate the relative importance of different constructs in adoption vs. intention to adopt internet banking. This study has presented an integrated framework of first order relations between 14 constructs from different theories. The results were extremely consistent and show a clear theoretical and empirical direction towards mobile banking consumer behavior for users. Other studies have applied some of the proposed theories in the mobile banking context (Luarn and Lin, 2005; Pedersen, 2005; Suoranta and Mattila, 2004), but none have applied all theories in the same framework, assessing the individual parts of a consumer s technology adoption process. Constructs Path coefficients (PC) LV index values f 2 Q 2 Table VIII. Path analysis mobile banking users Perceived ease of use! Attitude 0.505 5.161 0.270 0.591 Relative advantage! Attitude 0.458 4.447 0.135 0.582 Visibility! Attitude 0.454 4.316 0.135 0.562 Technology facilitating condition! Perceived behavioral control 0.416 4.526 0.454 0.558 Self-efficacy! Perceived behavioral control 0.373 5.209 0.642 0.724 Compatibility! Attitude 0.326 4.844 0.041 0.617 Perceived behavioral control! Intention 0.229 4.962 0.074 0.464 Resource facilitating condition! Perceived behavioral control 0.215 4.572 0.074 0.568 Subjective norm! Intention 0.178 2.727 0.080 0.657 Attitude! Intention 0.156 4.571 0.037 0.528 Results demonstrability! Attitude 0.147 4.563 0.016 0.648 Trialability! Attitude 20.208 4.818 0.025 0.589 Image! Attitude 20.267 2.993 0.074 0.629
Mobile banking 405 Figure 3. Mobile banking integrated adoption framework users The proposed framework showed a good R 2 for the intention construct, obtaining almost 69 percent explained variance when applied to the mobile banking non-user group. Technology adoption studies normally result in explaining about 40 percent of the intention to use a specific technology variance (Venkatesh et al., 2003). However, the proposed framework obtained only 22 percent explained variance when applied to mobile banking users. In addition, only three path coefficients had medium and large effect sizes in this case, suggesting that the framework and its LV need to be reviewed in order to predict intention to continue using an innovation, in this case mobile banking. In terms of a marketing strategy that seeks to achieve potential adopters, the most important factors that influence the consumer to adopt mobile banking are: Attitude! Intention (0.657); Technology facilitating control! Perceived behavioral control (0.574); Self-efficacy! Perceived behavioral control (0.503); Relative advantage! Attitude (0.459); Compatibility! Attitude (0.342) and Subjective norm! Intention (0.141). Based on these results, an understanding of consumer behavior towards mobile banking requires concepts from different theories. To achieve successful marketing strategies it is recommended that decision-makers should base their action plans on the relevant factors that impact the consumer.
IJBM 28,5 406 Conclusions This study has two key objectives - to propose a new integrated framework for studying innovation adoption and to apply this framework within the mobile banking context in order to investigate the factors that influence the adoption of mobile banking in Brazil. The study collaborates with many factors that influence the adoption of innovations. These factors help provide a deeper understanding of consumers innovation adoption process. The study proved to be very consistent in terms of the results obtained for the 13 hypothesized relations, highlighting, on the one hand, a theoretical direction that is more integrative of the adoption of new technologies, supported by five different existing theories. Furthermore, the proposed framework is more comprehensive than those preceding it and therefore is more suitable for a multitude of situations and contexts. Variables considered in this study can provide important insights for the development of new mobile services. One example is compatibility with life-style; this is one of the most important factors to be considered by a manager when launching new mobile services. Despite the insightful results, the study has some limitations that need to be considered when analyzing the results. The primary survey was conducted in Brazil via online questionnaires. So the results cannot be extrapolated to a global reality and even in Brazil, the sample only considered those who have access to the internet (due to the online questionnaire); the results are not a picture of the population s perceptions and behaviors. The proposed framework needs further studies to understand the factors that will motivate a customer to continue using an innovation. Although the study was not able to statistically compare the results between mobile banking users and non-users, it shed some light on possible differences between these groups. However, the proposed framework offered few contributions when it comes to understanding mobile banking users, since it resulted in a small or null effect size in most of the path coefficients. In addition, some aspects in the model, such as image and trialability, did not work as expected. It is suggested that future studies focus more on understanding why these factors, normally important in information technology adoption, were not significant within the mobile banking context. Studies that look at the adoption process by comparing fixed and mobile technology (i.e. internet banking and mobile banking), might be interesting for shedding light on the differences between the adoption process of fixed and mobile technologies. References Ajzen, I. (1985), From intentions to actions: a theory of planned behavior, in Kuhl, J. and Beckman, J. (Eds), Action-control: From Cognition to Behavior, Springer, Berlin, pp. 11-39. Ajzen, I. (1991), The theory of planned behaviour, Organizational Behavior and Human Decision Processes, Vol. 50 No. 2, pp. 179-211. Ajzen, I. and Fishbein, M. (1969), The prediction of behavioral intentions in a choice situation, Journal of Experimental Social Psychology, Vol. 5, pp. 400-16. Ajzen, I. and Fishbein, M. (1970), The prediction of behaviour from attitudinal and normative variables, Journal of Experimental Social Psychology, Vol. 6, pp. 466-87. Ajzen, I. and Fishbein, M. (1980), Understanding Attitudes and Predicting Social Behavior, Prentice-Hall, Englewood Cliffs, NJ.