THE IMPACT OF FLOW ON ONLINE CONSUMER BEHAVIOR

SANG M. LEE University of Nebraska-Lincoln Lincoln, NE 68588-0491

ABSTRACT Previous research has acknowledged flow as a useful construct for explaining online consumer behavior. However, there is dearth of knowledge about what dimensions of flow and how they actually influence online consumer behavior as flow is difficult to conceptualize and measure. This research examines flow and its effects on online consumer behavior in a unified model which draws upon theory of planned behavior (TPB). The four important dimensions of flow (concentration, enjoyment, time distortion, telepresence) are explored in terms of their antecedent effects on online consumer behavior. Results of this empirical study show that flow influences online consumer behavior through several important latent constructs. Findings of this research not only extend the existing knowledge of flow and its antecedent effects on online consumer behavior but also provide new insights into how flow can be conceptualized and studied in the e-commerce setting. Keywords: online consumer behavior, flow, concentration, enjoyment, time distortion, telepresence, TPB, TAM 1. INTRODUCTION The study of online consumer behavior has been one of the major research agendas in both information systems and marketing science with the rapidly growing e-business. However, the study is still at its infant stage and there is very limited knowledge about how online consumer behavior is really shaped. There are two major barriers to the study. First, online consumer behavior is a complicated socio-technical phenomenon that involves many interdependent factors that together shape online consumer behavior. The factors come from various disciplines including marketing, information systems, psychology, sociology, economics, etc. Therefore, online consumer behavior is a highly complex nonlinear phenomenon and no single model or theory can fully describe and explain it. Second, online consumer behavior is still under development with the advancing information technology and the rapidly changing social environment. This study extends theory of planned behavior (TPB) [2] by including several important constructs that have been identified to have important influence on online consumer behavior. TPB is a wellestablished general theory of social psychology in explaining and predicting human behavior. It takes into account the effects of the attitude toward behavior, social influence and personal behavioral control in the consumer decision process. Online consumers are both consumers and information systems users [25]. Although traditional consumer behavior is well described by marketing and economic theories, there is Received: October 12, 2009 Summer 2010

LIqIANG CHEN University of Nebraska-Lincoln Lincoln, NE 68588-0491

overwhelming evidence which suggests that information systemsrelated variables have become at least as important as traditional marketing factors in predicting online consumer behavior [14, 30]. Therefore, it is more appropriate to view e-commerce as an information system phenomenon where the consumer interacts with a complex information system than viewing it as a marketing issue influenced by information systems [25]. This study follows this research belief and examines online consumer behavior from the information systems user perspective. Two important beliefs from the technology acceptance model (TAM) [12] perceived ease-of-use (PEOU) and perceived usefulness (PU) were thus introduced into the TPB-based model as external attitudinal beliefs of information systems. However, a shortcoming of TAM is that it looks at the information system use from a utilitarian perspective which is not sufficient when considering contexts in which technology users are free to select an information system based on their feelings [34]. Recent studies indicate that ignoring the hedonic aspects of information systems often gives an incomplete picture or even inconsistent findings in the study of online consumer behavior. Flow as an important hedonic construct has been widely recognized to have a major impact on information system user behavior [10]. According to Csikszentmihalyi [10, 11], flow is the crucial component of enjoyment, and it represents the holistic sensation that people feel when they act with total involvement. Flow has also been acknowledged to be a useful variable for explaining online consumer behavior [21, 25]. Siekpe [33] indicated that the flow construct has been proposed as important for understanding the nature of user online experience. Researchers assert that the benefits of flow online include increased learning, exploratory and positive behavior, positive subjective experience, and perceived sense of control over their interaction. Nevertheless, flow is ill-defined because of the numerous ways it is conceptualized, operationalized, and measured [25]. There have been over ten different ways of conceptualizing flow. Novak et al. [29] found that on average scholars considered only four dimensions or constructs of flow in pervious studies. For example, Koufaris [25] applied three dimensions of flow (enjoyment, control, concentration) in his online consumer behavior model. Such multidimensional approach to flow measurement in the research model has been widely adopted in the online environment. One advantage of using multidimensional measurements of flow is that it allows statistical tests of whether the constituent constructs should be viewed as part of a higherorder factor measuring flow, or as antecedents or consequences of this factor [22]. Most recently, Hoffman and Novak [22] recommended that the multidimensional approach should be used whenever possible. Accepted: January 28, 2010 1

Revised: January 6, 2010 Journal of Computer Information Systems

This study views flow as one of the most important antecedents that influence online consumer behavior and conceptualizes flow with four major constructs based on previous studies. The four constructs are concentration, enjoyment, time distortion, and telepresence. In this study, these four flow constructs were introduced into the TBP-based model as external behavioral beliefs in explaining the actual online consumer behavior. 2. RESEARCH MODEL AND HYPOTHESES This study takes a positivist view of online consumer behavior. TPB plays a theoretical foundation of the research model in the study. According to Ajzen [2], TPB delineates specific salient beliefs that influence given behavioral perceptions as well as subsequent actual behavior; there are three major behavioral perceptions which are personal attitude, social influence, and personal behavioral control that finally determine actual behavior. Figure 1 shows the TPB-based research model in this study. TPB introduces perceived behavioral control (PBC) to take into account the effect of behavioral control. PBC refers to an individuals perception of how easy or difficult it is to perform a behavior and it reflects beliefs regarding access to resources and opportunities required to facilitate a behavior [2]. PBC not only positively influences intention along with attitude and subjective norm (SN), but also positively determines the final behavior with intention together [2]. Such dual role of PBC has been empirically supported by Mathieson [28] and Taylor and Todd [37]. Neglecting PBC and relying on simpler models may lead the study of online consumer behavior to be incomplete and potentially misleading [30]. This is because online consumers face several new constraints, such as the impersonal nature of the online environment, the extensive use of IT, and the uncertainty of the open Internet infrastructure. These issues call for the inclusion of perceived behavioral control (PBC) in e-commerce adoption models [30]. It is noteworthy that PBCs direct influence on intention to purchase in TPB has been acknowledged to have two distinct dimensions: self-efficacy and controllability [31]. Pavlou and Fygenson [31] conceptualized self-efficacy as consumers judgments of their own capabilities to conduct online consumption and controllability as consumers judgments about the availability of resources and opportunities to perform such consumption. They further viewed PBC as a second-order construct which was formed by the first-order dimensions of self-efficacy and controllability in their extended TPB model. Such relationship (the higher order construct is formed by the lower order factors) is typically a formative relationship or structure. With a formative relationship, PBC can be thoroughly predicted through its underlying dimensions and their respective antecedents, while maintaining a parsimonious view of PBC [31]. This postulation was further supported by a recent study of information technology continuance conducted by Bhattacherjee et al. [5]. They indicated that the two dimensions of PBC, self-efficacy and facilitating conditions (controllability), were linking to IT continuance [5]. This study adopted the formative structure of PBC in the model shown in Figure 1. The external beliefs shape PBC through their direct influence on self-efficacy and controllability. In the online environment, usage of information systems is one of the most important external behavioral beliefs that influence personal attitude and personal behavioral control [12, 31]. As 2

such, online purchase intentions should be explained in part by TAM [14]. TAM posits that PU and PEOU influence behavioral intention through attitude [12, 37]. The behavioral effects of PU and PEOU on online consumer behavior have been well confirmed by the majority of studies [14, 25, 30]. PEOU has also shown a significant impact on PU [12, 37, 38]. Beside its behavioral effect on intention, PEOU is also viewed as a control belief that facilitates a behavior with a lower personal effort [26]. Davis [12] argued that self-efficacy was one of the means by which PEOU influences behavior. Recently, Pavlou and Fygenson [31] further argued that in the e-commerce setting, an easy to use website was likely to increase the consumers confidence and remove the cognitive impediments during conducting online activities. Accordingly, it is hypothesized that perceived usefulness (PU) directly influences the attitude toward purchasing and that perceived ease-of-use (PEOU) directly impacts on the attitude toward purchasing and perceived behavioral control (PBC). The TAM related hypotheses are shown in Figure 1. Online information systems provide the consumer with a virtual environment in which flow is generated. In such virtual environment, flow has shown to significantly influence both attitudes and intentions [1, 20, 23, 25, 32]. In order to be introduced into TPB, flow must be conceptualized and operationalized into several measurable constructs. Ghani et al. [16] indicated that enjoyment and concentration were the two most important dimensions of flow. Since then, these constructs have been empirically tested in several studies [6, 7, 20, 25, 29, 33]. Novak et al. [29] used a structural modeling approach to empirically examine the effects of various flow constructs on customer online behavior. Their findings indicated that telepresence and time distortion significantly influenced behavior but the effect of concentration was not supported. In this study, we revisit the behavioral effects from four important constructs of flow concentration, enjoyment, telepresence, and time distortion in a TPB-based model. Figure 1 shows these behavioral effects of flow and the corresponding hypotheses. In addition to the behavioral effects, flow also increases PBC [20, 23] and PEOU [1]. Recently, Sun and Zhang [36] empirically examined the causal relationship between PEOU and perceived enjoyment. They found that perceived enjoyment as typical flow construct had a significant impact on PEOU. We believe that like enjoyment, concentration helps enhance the consumers confidence and controlability and increases PEOU of online information systems. Similarly, we also believe that telepresence and time distortion impact PBC and PEOU since both increase the consumers sense of virtual environment in which he/she feels to have more control and manipulation possibilities of online information systems [33]. The flow effects on PBC and PEOU are shown in Figure 1. According to the research model presented in Figure 1, the following hypotheses are proposed: H1a: Concentration positively influences attitude toward online purchasing. H1b: Enjoyment positively influences attitude toward online purchasing. H1c: Time distortion positively influences attitude toward online purchasing. H1d: Telepresence positively influences attitude toward online purchasing. H2a: Concentration positively influences intention to purchase online. Summer 2010

Journal of Computer Information Systems

subjects are shown in Table 1. The subjects were all experienced online consumers in terms of years of Internet experience and times of online purchasing. The measures of various constructs were derived from previous researches. Appendix A lists a complete instrument items that measure the constructs in the model. All measurements used 7-point Likert scales except for a Yes/No question for actual behavior. A pilot study with 30 students was performed to test the validity and reliability of the survey instrument and to provide preliminary results of data analysis. The feedback from the pilot study was used to refine the instrument for clarity and completeness. Path analysis is one of the most important statistical tools to specify and test a structural model that reflects prior hypotheses about causal relationships among variables [24]. There are two distinct types of path analysis techniques Figure 1. A TPB-Based Model for Online Consumer Behavior in model testing, covariance-based structured equation modeling (SEM) and component-based partial-least-squares (PLS). PLS is different from SEM in terms of H2b: Enjoyment positively influences intention to the objective of analysis, statistical assumptions, and nature of the purchase online. fit statistics [15]. SEM is often used to test a priori specified or a H2c: Time distortion positively influences intention to sound theory based model using sample derived estimates against purchase online. population. PLS, on the other hand, is more suited for predictive H2d: Telepresence positively influences intention to applications and theory building [15]. PLS is recommended in purchase online. an early stage of theoretical development to validate exploratory H3a: Concentration positively influences self-efficacy models and therefore PLS usually assists researchers who focus and controllability during online purchasing. on the explanation of endogenous constructs [19]. H3b: Enjoyment positively influences self-efficacy and Whether using SEM or PLS depends the research objectives controllability during online purchasing. and research data type and availability. According to Henseler H3c: Time distortion positively influences self-efficacy et al. [19], PLS is recommended in the following research and controllability during online purchasing. situations: H3d: Telepresence positively influences self-efficacy and controllability during online purchasing. The sample size is small in regards to the number of latent H4a: Concentration positively influences PEOU. variables. H4b: Enjoyment positively influences PEOU. The model is complex and has many latent and manifest H4c: Time distortion positively influences PEOU. variables. H4d: Telepresence positively influences PEOU. The model has less stringent assumptions about the distribution of variables and error terms. Although this study was primarily interested in how various The model has both reflective and formative variables. flow dimensions impact online consumer behavior, the effects of PEOU, PU, and various TPB-related variables would also be On the other hand, SEM usually requires a much larger sample empirically tested. However, those non-flow variables effects size. According to Kline [24], a desired sample size would be 20 were not listed as hypotheses in the study. times free parameters, or that a realistic target would be 10 times free parameters. For example, the research model in this study 3. RESEARCH METHODOLOGY consists of over 30 free parameters, and thus the required sample size by SEM could reach 300 or more. In addition, SEM only An online survey methodology was adopted. The questionnaire works with reflective indicators or structures [8, 9, 19]. was administered to college students across different majors Compared with SEM, PLS works better with a small sample at a large state university in the United States. There were 288 size and supports both reflective and formative structures [19]. respondents and their complete questionnaires were used for data The small sample requirement in PLS results from the fact that analysis and model testing. College students represent a major PLS conducts principal components regression only on those online consumer group. Most of them are skillful and innovative closely connected latent variables and seeks for local optimization in using the Internet. College students are also a homogenous with fewer observed variables or indicators involved [8, 9]. In this group and many exogenous factors can be easily controlled when study, the research model is complex and it includes both reflective using them as a sample. By using college students a sample, and formative structures. In addition, the objective of this study is we aim to find the online consumer behavior of such consumer to examine the effects of several endogenous constructs including group in market segmentation. To meet the random sampling flow and TAM beliefs, and thus the study is more explanatory and assumption required by structural modeling, we collected the prediction-oriented than theory building. Therefore, PLS is a well data from randomly selected subjects. The demographics of the Summer 2010 Journal of Computer Information Systems 3

Table 1. Sample Profile Variable Gender: Male Female Age: 19-24 25 and above Years of Internet Experience: At least 4 years 0-3 years Weekly Hours of Internet Use: At least 5 hours less than 5 hours Online Purchase Frequency: At least once every 6 months 0-once a year suited statistic tool in this study and was employed to analyze a complete survey dataset. In path analysis and hypothesis testing, PLS applies either a jackknife or a bootstrap approach to estimate the significance (t-values) of the paths. Neither of these PLS significance estimation methods require parametric assumptions [8]. This study used the bootstrap approach with 500 re-samples to test the significance of path and hypotheses in the model. Efron and Tibshirani [13] suggested that 500 resamples are sufficient for the general standard bootstrap method in most cases. Similarly, Manly [27] indicated that 200 re-samples generally gave a relatively small error margin in bootstrap estimation. In general, the number of re-sample in the bootstrap approach is determined by the levels of significance and confidence intervals (CI). Thus, a larger number of re-sample will be required by a high significant level and/or larger CI. The literature review showed many publications adopted 200-500 re-samples as a rule of thumb without giving reasons. 4. ANALYSIS AND RESULTS 4.1 Measurement Validity SmartPLS software (http://smartpls. com) was used to perform both instrument validation and structural path modeling. The measurement was subjected to reliability and validity analyses prior to the path analysis and hypothesis test. Both Cronbachs alpha and composite reliability were used to assess 4 # of Subjects 159 129 240 48 279 9 263 25 Percentage (%) 55.2 44.8 83.4 16.6 96.9 3.1 91.3 8.7

Table 2. Results of Reliability Cronbachs and Composite Reliability Construct Attitude Behavior Concentration Controllability Enjoyment Intention PEOU PU Self-Efficacy Subject Norm Telepresence Time Distortion # of Indicators 2 1 4 2 4 2 4 4 2 2 5 2 Cronbachs 0.909 1.000 0.905 0.766 0.925 0.970 0.914 0.936 0.970 0.877 0.946 0.836 Composite Reliability 0.957 1.000 0.933 0.895 0.945 0.985 0.939 0.954 0.985 0.942 0.955 0.907

245 43

85.1 14.9 Attitude ATTI_1 ATTI_2 BEHA CNCN_1 CNCN_2 CNCN_3 CNCN_4 CNTR_1 CNTR_2 ENJY_1 ENJY_2 ENJY_3 ENJY_4 INTE_1 INTE_2 PEOU_1 PEOU_2 PEOU_3 PEOU_4 PUSE_1 PUSE_2 PUSE_3 PUSE_4 SEFF_1 SEFF_2 SUBJ_1 SUBJ_2 TDST_1 TDST_2 TELE_1 TELE_2 TELE_3 TELE_4 TELE_5 0.959 0.956 0.617 0.317 0.355 0.313 0.245 0.602 0.546 0.393 0.470 0.293 0.342 0.642 0.641 0.657 0.606 0.609 0.657 0.393 0.413 0.382 0.364 0.574 0.590 0.655 0.557 0.211 0.079 0.048 -0.050 -0.045 -0.051 -0.114

Table 3(1). Results of Validity Cross Loadings Behavior 0.592 0.590 1.000 0.207 0.267 0.212 0.183 0.421 0.393 0.251 0.310 0.174 0.207 0.517 0.514 0.464 0.450 0.404 0.429 0.248 0.280 0.248 0.212 0.426 0.451 0.406 0.342 0.127 0.076 0.012 -0.087 -0.039 -0.021 -0.071 Concentration 0.322 0.360 0.252 0.867 0.891 0.908 0.856 0.312 0.321 0.624 0.626 0.695 0.606 0.244 0.235 0.311 0.252 0.340 0.347 0.252 0.292 0.339 0.288 0.302 0.296 0.355 0.284 0.507 0.400 0.396 0.280 0.354 0.351 0.296 Control 0.622 0.600 0.453 0.244 0.411 0.352 0.186 0.904 0.896 0.308 0.402 0.183 0.271 0.551 0.526 0.667 0.628 0.589 0.623 0.292 0.348 0.309 0.292 0.715 0.724 0.606 0.557 0.146 0.009 -0.078 -0.181 -0.120 -0.107 -0.181 Enjoy 0.381 0.441 0.272 0.663 0.550 0.618 0.665 0.313 0.297 0.903 0.929 0.878 0.895 0.280 0.277 0.376 0.338 0.441 0.448 0.357 0.370 0.418 0.347 0.319 0.335 0.411 0.330 0.557 0.418 0.353 0.215 0.282 0.248 0.226 Intention 0.630 0.616 0.523 0.226 0.277 0.189 0.132 0.542 0.439 0.292 0.323 0.158 0.196 0.985 0.985 0.518 0.508 0.406 0.447 0.229 0.246 0.204 0.204 0.557 0.542 0.513 0.459 0.121 0.024 -0.037 -0.140 -0.106 -0.083 -0.144

Journal of Computer Information Systems

Summer 2010

the reliability. The accepted values for both Cronbachs alpha and composite reliability are 0.70 or higher. Table 2 shows the SmartPLS output of reliability testing. All Cronbachs alpha and composite reliability values listed in Table 2 are greater than 0.70, indicating the measurement instrument of the study is reliable. Two important construct validities convergent validity and discriminant validity were accessed in the study. According to Chin [8], both convergent and discriminant validities can be inferred if the PLS indicators do the following: (1) load much higher on their measured construct than on other constructs, that is, the own-loadings are higher than the cross-loadings, and (2) when the square root of each constructs average variance extracted (AVE) is larger than its correlations with other Table 3(2). Results of Validity Cross Loadings PEOU ATTI_1 ATTI_2 BEHA CNCN_1 CNCN_2 CNCN_3 CNCN_4 CNTR_1 CNTR_2 ENJY_1 ENJY_2 ENJY_3 ENJY_4 INTE_1 INTE_2 PEOU_1 PEOU_2 PEOU_3 PEOU_4 PUSE_1 PUSE_2 PUSE_3 PUSE_4 SEFF_1 SEFF_2 SUBJ_1 SUBJ_2 TDST_1 TDST_2 TELE_1 TELE_2 TELE_3 TELE_4 TELE_5 0.689 0.671 0.491 0.283 0.378 0.315 0.228 0.657 0.610 0.409 0.497 0.301 0.360 0.516 0.525 0.893 0.898 0.874 0.901 0.448 0.501 0.445 0.438 0.739 0.739 0.542 0.517 0.248 0.087 0.032 -0.060 0.014 -0.047 -0.086 PU 0.413 0.399 0.271 0.294 0.270 0.301 0.262 0.325 0.286 0.411 0.370 0.340 0.338 0.247 0.229 0.426 0.381 0.470 0.511 0.880 0.918 0.949 0.918 0.366 0.364 0.374 0.388 0.262 0.256 0.219 0.194 0.176 0.193 0.150 SelfEfficacy 0.573 0.558 0.445 0.204 0.388 0.283 0.142 0.671 0.644 0.348 0.389 0.164 0.228 0.551 0.548 0.760 0.687 0.584 0.634 0.341 0.376 0.334 0.301 0.985 0.985 0.562 0.463 0.139 -0.023 -0.080 -0.218 -0.148 -0.174 -0.250 Subject Norm 0.632 0.603 0.398 0.272 0.375 0.320 0.194 0.609 0.501 0.323 0.429 0.288 0.355 0.533 0.485 0.506 0.493 0.470 0.532 0.320 0.403 0.362 0.391 0.525 0.550 0.950 0.937 0.222 0.090 0.052 0.018 0.029 0.054 -0.023 Time Distortion 0.171 0.190 0.121 0.518 0.396 0.380 0.525 0.121 0.091 0.522 0.472 0.531 0.485 0.104 0.098 0.198 0.134 0.224 0.228 0.253 0.233 0.252 0.275 0.105 0.101 0.228 0.145 0.983 0.834 0.631 0.444 0.447 0.434 0.374

constructs. Convergent validity describes the degree to which a measure is correlated with other measures in a single variable measurement. Discriminant validity refers to the degree to which the measurement for one variable does not correlate with the measurement for another variable. Table 3 represents the item loadings on their measured constructs. All items are well loaded on their constructs, that is, their own (on their measured construct) loadings (in bold font in Table 3) are much higher than the cross loadings (on other constructs). Table 4 shows the AVE values for all constructs. The accepted AVE should be above 0.5 in order to achieve convergent and discriminant validities. The testing results of both cross loadings and AVEs suggest that all construct measurements have adequate convergent and discriminant validities. 4.2 PLS Path Modeling and Hypotheses Testing Figure 2 shows the path coefficients and their corresponding t-values. The bootstrap approach (500 re-samples) was used to test the significance of path and hypothesis in the path model. One tail t-test was used since all hypotheses are directional in the study. According to one tail t-test (df = 500), the 95% significance level or p<0.05 requires t-value>1.645 and the 99% significance level or p<0.01 requires t-value>2.340. When df >100, the t-test is actually very close to z-test. In Figure 2, path coefficients and t-values are shown only for the significant paths which are represented in solid lines, and those non-significant paths are not drawn. The empirical data testing results support many important hypotheses listed in section 2. The supported hypotheses are H1a (concentration -> attitude), H1d (telepresence -> attitude), H3a (concentration -> controllability & self-efficacy), H3d (telepresence -> controllability & self-efficacy), H4a (concentration -> PEOU), H4b (enjoyment -> PEOU), and Table 4. Results of Validity AVE AVE Attitude Behavior Concentration Control Enjoy Intention PEOU PU Self-Efficacy Subject Norm Telepresence Time Distortion 5 0.917 1.000 0.776 0.810 0.812 0.971 0.795 0.840 0.971 0.890 0.810 0.831

Telepresence -0.081 -0.061 -0.063 0.354 0.212 0.275 0.405 -0.153 -0.146 0.196 0.168 0.362 0.277 -0.118 -0.144 -0.121 -0.080 0.015 0.003 0.127 0.154 0.204 0.216 -0.227 -0.208 -0.021 0.053 0.394 0.547 0.788 0.912 0.916 0.938 0.938

Summer 2010

Journal of Computer Information Systems

Figure 2. PLS Path Model Note: Only solid lines with significant path coefficient and t-value are shown. For H3a and H3d, the first pair of path coefficient-c and t-value is for controllability; the second pair is for self-efficacy. *p<0.05, **p<0.01

H4d (telepresence -> PEOU). In summary, concentration and telepresence were supported to have impacts on the attitude toward purchasing, PBC and PEOU; enjoyment was supported to only influence PEOU. In addition to testing the proposed hypotheses in the model, this study re-examined the TPB propositions and the effects of TAM beliefs PEOU and PU. All TPB propositions were highly supported at the p<0.01 significance level except for subject norms effect on intention to purchase which was not supported by the study. PU was supported to have impact on the attitude toward purchasing at the p<0.05 significance level. PEOU was strongly supported to influence the attitude toward purchasing and PBC at the p<0.01 significance level. 5. DISCUSSION OF FINDINGS AND IMPLICATIONS 5.1 Research Findings Concentration was significantly supported to have a positive influence on attitude (H1a). This finding is in line with Koufais [25] conclusions concentration has been a significant correlate or measure of flow; it can also play a role in online consumer behavior and be critical for completing their purchases efficiently. Concentration helps consumers reach their online purchasing goals [25]. In addition, concentration was found to have a positive impact on PCB through controllability and selfefficacy (H3a). This finding confirmed Hoffman and Novaks [20] proposition that flow increases PBC. In such relationship, PBC is the consequence of concentration [29]. The findings also supported that concentration positively influenced PEOU (H4a). This finding further verified Agarwal and Karahannas [1] proposition that flow is a proximal antecedent of PEOU. Therefore, our findings indicated that the antecedent effects of concentration on online consumer behavior were actually achieved and moderated by attitude, PBC, and PEOU. Similarly, the antecedent effect of telepresence on the attitude toward purchasing (H1d) was also significantly supported by the findings. But this relation was negative, indicating that telepresence decreases the attitude toward purchasing. 6

Telepresence was also found to have a negative influence on PBC (H3d) and PEOU (H4d). The negative effects of telepresence might not be hard to understand. Most online consumers usually have well-established behaviors from their previous experiences in the physical environment. When they were presented with a virtual environment which is measured by telepresence, they often felt lacking social connection, trust, or help. Therefore, such virtual environment could cause consumers to feel highly uncertain or even nervous so that their behavioral attitudes toward purchasing would decrease. Actually, telepresence describes such nervous feelings in a virtual environment as to the extent to which one feels present in the mediated environment, rather than in the immediate physical environment [35]. Interestingly, flow has been acknowledged to play a dual role in a cognitive behavioral process. Flow represents itself as flow experience sometimes [16] and as an antecedent to flow in other times [29]. We argue that during the online purchasing process, concentration more likely acts as an antecedent to flow which influences attitude, PBC, and PEOU. This argument is in line with Koufaris [25] findings because concentration helps consumers reach their goals. Similarly, telepresence is often considered more as an antecedent to flow than as the flow experience itself [20, 29]. With these in mind, we recommend that e-commerce websites should reduce telepresence as much as possible to remove consumers nervousness in the online environment; meanwhile, they should also remove advertisements and purchasing-unrelated contents to enhance consumers concentrations. Enjoyment is another flow construct that was empirically supported to shape online consumer behavior. Such antecedent effect was realized only through PEOU. There was no empirical support that enjoyment has any effect on PBC or attitude in this study. There are two plausible explanations for this finding. First, the participants might not have any enjoyment or joyfulness during the online purchasing process because online shopping is a simple and easy task. This explanation was partially supported by the pilot study in which some subjects indicted online shopping was easy and no fun at all. Today, most e-commerce websites provide easeto-use interfaces for searching and purchasing purposes and thus the online purchasing processes is quite easy and straightforward. Summer 2010

Journal of Computer Information Systems

The second explanation is that enjoyment might not be recalled in this study because a retrospective method was used to measure the subjects past flow experiences without specifying which online shopping activity they would recall/report. Flow exists in a process [10] and is context-dependent and thus it could be easily missed out in the context-free measure instrument. The empirical support of enjoyment effect on PEOU implies that information systems play a critical role in determining online consumer behavior. In other words, online consumer may obtain enjoyment from using information systems instead of from the shopping task. Such implication should be emphasized in all the studies of information system phenomena because information systems act as a direct interface for behavioral formation. The antecedent effects of time distortion were not empirically supported by this study. This could be explained by the fact that when people have equal experiences from the real world and the virtual world, they no long see the time distortion happening in the virtual world. Today, many online consumers, especially college students, are skillful online and offline consumers and thus they have equivalent experiences both online and offline. This explanation was also partially supported by the pilot study interview, that is, some subjects reported that they did not feel time passing faster while they were doing online shopping because the shopping process was easy and just like what they did on the Internet every day. In summary, this study found three important dimensions of flow (concentration, telepresence, and enjoyment) influence online consumer behavior through three important latent constructs (attitude, PBC and PEOU) of information systems. Enjoyment was found to have a direct impact only on PEOU which represents the critical determining role of information systems in online consumer behavior. Concentration and telperesence shape actual online consumer behavior through two behavioral dimensions: consumer dimension (attitude toward purchasing and PBC) and information systems dimension (PEOU). Both concentration and enjoyment help improve online consumer behavior, but telepresence inhibits such behavior due to concern and nervousness in the virtual environment. In addition to testing the flow effect on online consumer behavior through various intermediate latent constructs in the TPB-based model, the findings of the study also supported and confirmed the direct effects of PEOU and PU on attitude as shown in Figure 2. PEOU was also supported to directly impact PBC through both controllability and self-efficacy. The subject norm which measures social influence on consumer behavior was not supported by this study. This is consistent with Pavlou and Fygensons [31] findings. Such finding may indicate online consumer behavior lacks social influence from their friends, relatives, colleagues, etc. However, it is noteworthy that such situation could be greatly changed by online customer feedback/ ranking systems and social networks. What other experienced customers report as the feedback of Amazon.com may greatly influence new customers purchase decision making. 5.2 Implications for Academics This study adopted the positivist paradigm by taking into account the effects of flow and TAM beliefs in the TPB-based model. The study further confirmed and expanded earlier research in the area and enriched the existing knowledge of flow and its effects on online consumer behavior. Specifically, the findings provided the empirical supports on the effects of three flow Summer 2010

constructs (concentration, enjoyment, and telepresence) on online consumer behavior. The study found telepresence had a negative impact on online consumer behavior and such finding is reported for the first time in the e-commerce setting. PEOU plays an important role in transferring and mediating flow effects on online consumer behavior. This finding further confirms the previous postulation that IT-related factors may play a more critical role in predicting online consumer behavior [14, 25, 31]. Similarly, attitude and PBC are other two important latent constructs that moderate the flow effects on actual consumer behavior. This research also provided further empirical evidences showing that the TPB-based model is an effective research platform for studying various socio-technical behaviors beyond TPBs original constructs [31]. The empirical findings also indicated that there was great potential for the TPB-based model to manipulate many external beliefs without compromising the models goodness of fit by using an appropriate statistical technique partial least squares (PLS). This study thus represents another step forward as it explores flow and online consumer behavior. 5.3 Implications for E-commerce Practitioners The findings of this study provided a general guideline for both consumers and e-commerce vendors on how to improve their buying and selling practices. We believe that the findings would help online consumers better understand how concentration, enjoyment, and telepresence as external antecedent factors shape their online consumer behavior. For both consumers and e-commerce vendors, an ease-to-use information system plays a critical role in improving consumer behavior. The findings indicated that three important flow dimensions (concentration, enjoyment, and telepresence) shape actual consumer behavior through PEOU. Therefore, the online information systems should effectively provide the interface that helps consumers achieve concentration, enhance enjoyment, and reduce telepresence. Consumers would like to go to the website that provides ease-touse and more real world-like interface. If the e-commerce vendor has physical presentation, it should make its online stores a great deal like its physical store. This helps consumers establish the linkage between the website and the physical presence and thus increase trust and reduce concern and nervousness. Another way to overcome consumers concern and nervousness in the online environment is to provide user-friendly help and support systems. Such systems are especially helpful for less-skilled consumers to overcome various distractions and uncertainties during their online shopping processes. E-commerce vendors should improve their websites to increase consumers concentration by reducing advertisements and non-shopping related contents. 5.4 Limitations and Recommendations for Future Research This study has some methodological and theoretical limitations that can provide some opportunities for future research. This study used a retrospective or self-report method to obtain information about the participants past online shopping experiences. There are some major limitations of using such method. First, the data is not always assured of validity even though self-reports are more easily obtained [3]. Flow is transient information and is embedded in a particular process. Flow could be easily ignored or inaccurately reported by the participants if their flow experiences happened a few months ago. To overcome such problems in the self-report method, we recommend using a lab experiment to 7

Journal of Computer Information Systems

examine flow and its effects on online consumer behavior. A lab experiment usually records real-time information from subjects and the subjects behavioral information can be more accurately collected. Second, this study collected general information about participants past online purchasing experiences without specifying a particular e-commerce website or particular online purchasing activity. Such instrument design has a certain tradeoff. The positive side is that the data reflects consumers general experiences and thus the research findings may have high generalizability. The negative side is that flow may not be detected because flow is context dependent in terms of time, location and object. This limitation can be overcome by a lab experiment. Third, this study adopted the measurement instruments from prior research, mostly from the field of management information systems (MIS). Some instruments may not be well suited for the study of online consumer behavior or in the TPB-based model. For example, the measurement of enjoyment should be redeveloped rather than using the instrument developed by Ghani et al. [16]. This is because enjoyment has different meanings and implications in the online purchasing process from those in online entertaining, searching or surfing. Recently, in examining two general website usability instruments, Green and Pearson [18] found that the underlying theoretical constructs suggested by those instruments might not be sufficient to adequately measure website usability in an e-commerce setting. Therefore, the measurement of a construct should reflect its specific context. We recommend that future research should develop and use a context related measurement. Depending on research questions, the measurement of a construct must represent its unique institution. No universal measurement instrument currently exists. Lastly, we think this study missed an important factor emotion. Emotion is an important marketing and consumer behavioral factor, which has been studied in the traditional consumer behavior area for several decades. Bagozzi [4] asserted that emotion played an important role in determining consumer decision making. Koufaris [25] found that emotion existed in online shopping and had positive influence on online consumer behavior. By analyzing emotion (emphasizing the behavioral response caused by external stimuli) and flow (emphasizing the internal psychological state) together, we will better understand online consumer behavior. In conclusion, this study provides a valuable research model and empirical results in the area of online consumer behavior. Overcoming the limitations in future research will open a new avenue for the study of flow and online consumer behavior. This study examined only four dimensions of flow. Future research can introduce additional flow dimensions in the TPBbased model so that the interaction among various flow as well as their compound effects on online consumer behavior can be examined. We also recommend that future research reexamine the influence of subject norm which represents an increasingly important social influence from social networks and consumer online feedbacks. REFERENCES [1] Agarwal, R., and Karahanna, E., Time Flies When Youre Having Fun: Cognitive Absorption and Beliefs about Information Technology Usage, MIS Quarterly 24(4), 2000, pp. 665-694. [2] Ajzen, I., The Theory of Planned Behavior, Organizational 8

Behavior and Human Decision Processes 50, 1991, pp. 179211. [3] Ajzen, I., Constructing a TpB Questionnaire: Conceptual and Methodological Considerations, Working Paper, University of Massachusetts, Amherst. Retrieved in September 2002 from http://www-unix.oit.umass.edu/~aizen/pdf/tpb.mea surement.pdf. [4] Bagozzi, R. P., Goal-directed Behaviors in Marketing: Cognitive and Emotional Perspectives, Psychology & Mar keting 14(6), 1997, pp. 539-543. [5] Bhattacherjee, A., Perols, J., and Snaford C., Information Technology Continuance: A Theoretic Extension and Empirical Test, Journal of Computer Information Systems 49(1), 2008, pp. 17-26 [6] Chen, H., Wigand, R., and Nilan, M. S., Optimal Experience of Web Activities, Computers in Human Behavior, 15(5), 1999, pp. 585-608. [7] Childers, T., Carr, C., Peck, J., and Carson, S., Hedonic and Utilitarian Motivations for Online Retail Shopping Behavior, Journal of Retailing, 77(4), 2001, pp. 511535. [8] Chin, W. W., The Partial Least Squares Approach to Structural Equation Modeling, Modern Methods for Business Research, G. A. Marcoulides (ed.), Lawrence Erlbaum Associates, Mahwah, NJ, 1998, pp. 295-336. [9] Chin, W. W., Marcolin, B. L., and Newsted, P. R., A Partial Least Squares Latent Variable Modeling Approach for Measuring Interaction Effects: Results from a Monte Carlo Simulation Study and an Electronic-mail Emotion/Adoption Study, Information Systems Research 14(2), 2003, pp. 189217. [10] Csikszentmihalyi, M., Play and Intrinsic Rewards, Hum anistic Psychology 15(3), 1975, pp. 41-63. [11] Csikszentmihalyi, M., and Csikszentmihalyi, I. S., Optimal Experience: Psychological Studies of Flow in Consciousness, Cambridge University Press, Cambridge, 1988. [12] Davis, F. D., Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology, MIS Quarterly 13(3), 1989, pp. 319-340. [13] Efron, B., and Tibshirani, R. J., An Introduction to the Bootstrap, Chapman and Hall, New York, 1993. [14] Gefen, D., Karahanna, E., and Straub, D. W., Trust and TAM in Online Shopping: An Integrated Model, MIS Quarterly 27(1), 2003, pp. 51-90. [15] Gefen, D., Straub, D. W., and Boudreau, M.-C., Structural Equation Modeling and Regression: Guidelines for Research Practice, Communications of the Association for Information Systems 4(7), 2000, pp. 1-78. [16] Ghani, J. A., Supnick, R., and Rooney, P., The Experience of Flow in Computer-mediated and in Face-to-face Groups, Proceedings of 12th International Conference of Information Systems, New York, New York, 1991, pp. 229-237. [17] Ghani, J., Flow in Human Computer Interactions: Test of a Model, Human Factors in Information Systems: Emerging Theoretical Bases, Carey, J. (Ed.), Ablex Publishing Corp, New Jersey, 1995, pp 291-311. [18] Green, D. G., and Pearson, J. M., The Examination of Two Web Site Usability Instruments for Use in B2C E-Commerce Organization, Journal of Computer Information Systems 49(4), 2009, pp. 19-32. [19] Henseler, J., Ringle, C. M., and Sinkovics, R. R. The Use of Partial Least Squares Path Modeling in International Summer 2010

Journal of Computer Information Systems

Marketing, Advances in International Marketing 20, 2009, pp. 277-319. [20] Hoffman, D. L., and Novak, T. P., Marketing in Hypermedia Computer-mediated Environments: Conceptual Foundations, Journal of Marketing 60(3), 1996, pp. 50-69. [21] Hoffman, L. D., Novak, P. T., and Duhachek, A., The Influence of Goal-directed and Experiential Activities on Online Flow Experiences, Journal of Consumer Psychology 13(1/2), 2003, pp. 3-16. [22] Hoffman, D. L., and Novak, T. P., Flow Online: Lessons Learned and Future Prospects, Journal of Interactive Marketing 23(1), 2009, pp. 23-34. [23] Huang, M., Flow, Enduring and Situational Involvement in the Web Environment: A Tripartite Second-Order Examination, Psychology and Marketing 23(5), 2006, pp. 383-411. [24] Kline, R. B., Principles and Practice of Structural Equation Modeling, 2nd Edition, The Guilford Press, New York and London, 2005. [25] Koufaris, M., Applying the Technology Acceptance Model and Flow Theory to Online Consumer Behavior, Information Systems Research 13(2), 2002, pp. 205-223. [26] Lepper, M. R., Microcomputers in Education: Motivational and Social Issues, American Psychologist 40(1), 1985, pp. 1-18. [27] Manly, B. F. J., Randomization, Bootstrap and Monte Carlo Methods in Biology, 2nd Edition, Chapman and Hall, London, 1997. [28] Mathieson, K., Predicting User Intentions: Comparing the Technology Acceptance Model with the Theory of Planned Behavior, Information Systems Research 2(3), 1991, pp. 173-191. [29] Novak, T. P., Hoffman, D. L., and Yung, Y. F., Measuring the Customer Experience in Online Environments: A Structural

Modeling Approach, Marketing Science 19(1), 2000, pp. 22-42. [30] Pavlou, P. A., Consumer Acceptance of Electronic Commerce: Integrating Trust and Risk with the Technology Acceptance Model, International Journal of Electronic Commerce 7(3), 2003, pp. 69-103. [31] Pavlou, P. A., and Fygenson, M., Understanding and Predicting Electronic Commerce Adoption: An Extension of the Theory of Planned Behavior, MIS Quarterly 30(1), 2006, pp. 115-143. [32] Sanchez-Franco, M. J., Exploring the Influence of Gender on Web Usage via Partial Least Squares, Behavior and Information Technology 25(1), 2006, pp. 19-36. [33] Siekpe, J. S., An Examination of the Multidimensionality of Flow Construct in a Computer-Mediated environment, Journal of Electronic Commerce Research 6(1), 2005, pp. 31-43. [34] Sledgianowski, D., and Kulviwat, S., Using Social Network Sites: The Effects of Playfulness, Critical Mass and Trust in a Hedonic Context, Journal of Computer Information Systems 49(4), 2009, pp. 74-83. [35] Steuer, J., Defining Virtual Reality: Dimensions Determining Telepresence, Journal of Communication (42)4, 1991, pp. 73-93. [36] Sun, H., and Zhang, P., Causal Relationships between Perceived Enjoyment and Perceived Ease of Use: An Alternative Approach, Journal of Association for Information Systems (JAIS) 7(9), 2006, pp. 618-645. [37] Taylor, C. S., and Todd P. A., Understanding Information Technology Usage: A Test of Competing Models, Information Systems Research 6(3), 1995, pp. 144-176. [38] Venkatesh, V., and Davis, F. D., A Theoretical Extension of the Technology Acceptance Model: Four Longitudinal Field Studies, Management Science 46(2), 2000, pp. 186-204.

APPENDIx A: SURVEY ITEMS Online consumer behavior[19]: (yes/no) During the last three months, I purchased a product from an ecommerce website. Controllability [19]: (strongly disagree/strongly agree) All necessary resources for purchasing a product/service from an e-commerce website were accessible to me within the last three months. Purchasing a product/service from an e-commerce website was completely under my control within the last three months. Self-efficacy [19]: (strongly disagree/strongly agree) If I wanted to, I would be able to purchase a product/service from an e-commerce website. If I wanted to, I could purchase a product/service from an ecommerce website. Perceived Ease of Use [10]: (strongly disagree/strongly agree) It was easy to become skillful at the e-commerce website from which I made a purchase. Learning to operate the e-commerce website from which I made a purchase was easy. The e-commerce website from which I made a purchase was flexible to interact with. My interaction with the e-commerce website from which I made a purchase was clear and understandable. 9

Intention to purchase [19]: (strongly disagree/strongly agree) I intended to purchase a product/service from an e-commerce website before I made actual purchase. I planned to purchase a product/service from an e-commerce website before I made actual purchase Attitude toward online purchase [19]: (strongly disagree/strongly agree) For me, purchasing a product/service from an e-commerce website within the last three months was a good idea. For me, purchasing a product/service from an e-commerce website within the last three months was desirable. Subjective norm [19]: (strongly disagree/strongly agree) Most people who are important to me think that it is fine to purchase a product/service from an e-commerce website. Most people who are important to me would purchase a product/ service from an e-commerce website. Summer 2010

Journal of Computer Information Systems

Perceived Usefulness [10]: (strongly disagree/strongly agree) The e-commerce website from which I made a purchase improved my performance in online purchasing. The e-commerce website from which I made a purchase enabled me to online-purchase faster. The e-commerce website from which I made a purchase enhanced my effectiveness in online purchasing. The e-commerce website from which I made a purchase increases my productivity in online purchasing. Concentration [11]: (strongly disagree/strongly agree) I was absorbed intensely in the activity during my online shopping. My attention was focused on the activity during my online shopping. I concentrated fully on the activity during my online shopping. I was deeply engrossed in the activity during my online shopping. Enjoyment [11]: (strongly disagree/strongly agree) I found my online shopping interesting. I found my online shopping enjoyable. I found my online shopping exciting. I found my online shopping fun.

Telepresence [17]: (strongly disagree/strongly agree) I forget about my immediate surroundings when I used the ecommerce website. Using the e-commerce website often made me forget where I was. When I used the e-commerce website, I felt I was in a world created by the website I visited. After using the e-commerce website, I felt like I came back to the real world after a journey. When I used the e-commerce website, the world generated by the website I visited was more real for me than the real world. Time distortion [17]: (strongly disagree/strongly agree) Time seemed to go by very quickly when I used the e-commerce website. When I used the e-commerce website, I tended to lose track of time.

10

Journal of Computer Information Systems

Summer 2010

Copyright of Journal of Computer Information Systems is the property of International Association for Computer Information Systems and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.