The Vital Role of Trust in E-Commerce: A Meta-Analysis

Transcription

1 The Vital Role of Trust in E-Commerce: A Meta-Analysis Jun He College of Business, University of Michigan-Dearborn, Michigan, USA junhe@umich.edu Abstract- Trust plays a vital role in e-commerce to encourage desired transaction attitudes and behaviors from consumers. Research on trust in e-commerce has employed many different theoretical frameworks and tested a variety of variables, resulting in fragmented findings. This study attempts to synthesize previous empirical findings regarding antecedents and consequents of trust in e-commerce. A meta-analysis is conducted. 70 empirical studies have been identified, and 262 effect sizes are collected and metaanalyzed. Implications for future research are also discussed. Keywords- Meta-analysis; Trust; E-commerce; E-vendors; Trust Antecedents; Trust Effects I. INTRODUCTION Online transactions require customers to provide sensitive information in the absence of formal control mechanisms to monitor such information being appropriately used [32]. Also, compared with the traditional brick and mortar retailers, e- vendors lack some common means such as physical proximity and history of being in business to boost customer awareness and confidence [11]. Therefore, initiating, building, and maintaining trust among customers are deemed a key driver of success for e-vendors [3, 20, 23, 54]. E-commerce represents a complex blend of human actions and technological systems [20]. The formation and development of trust in such a rich context are shaped by various forces including economic, psychological, social, technological, and environmental factors. It is common for researchers to focus on certain aspects of e-commerce and investigate trust through special theoretical lens. As such, empirical research in the field has employed different perspectives and addressed a variety of variables, resulting in fragmented findings in the literature. This study attempts to systematically review empirical studies of trust in e-commerce. The purpose is to synthesize what we have examined, and make suggestions for future research. To serve the end, meta-analysis is selected as the primary research method. The remainder of the paper is organized as follows: first, the theoretical backgrounds of trust research in e- commerce are discussed; then, the meta-analytic procedure is described, and the results are reported. The paper ends with a discussion of the findings and the implications for future research. A. Conceptualization of Trust II. THEORETICAL BACKGROUND OF TRUST Trust has been studied in several disciplines including economics, sociology, marketing, and organizational behavior [11, 59]. Although there is no universally agreed definition of trust [48, 59], a generic typology proposed by Mayer et al. [45] has been widely adopted among trust researchers [59]. According to Mayer et al. [45], the construct of trust consists of three key dimensions of ability (trustor s perception of trustee s competencies and knowledge salient to the expected behavior), integrity (trustor s perception that the trustee will adhere to a set of principles that the trustor finds acceptable), and benevolence (the extent to which a trustee is believed to intend to do good to the trustor, beyond its own profit motive). Accordingly, IS researchers commonly define and operationalize trust in e-commerce as a collection of beliefs in an e- vendor s ability, integrity, and benevolence [11]. There are other aspects of trust being discussed in the literature, such as honesty [23], perceived risks [14], vendor legitimacy [67] and security [46, 67]. But the three dimensions of ability, integrity, and benevolence are the most adopted in the research of trust in e-commerce. B. The Formation and Development of Trust - Antecedents The research of trust in e-commerce has referenced different theoretical frameworks from economics, sociology, psychology, marketing, and organizational behavior. In this section, the antecedents of trust are discussed based on their theoretical grounds. 1) Personal Characteristics-based Trust Antecedents: Personal characteristics have been widely studied in the research of organizational behavior, especially in the area of group research [49]. An assumption of characteristics-oriented research is that people with similar characteristics have comparable

2 experiences and tend to behave in similar ways [77]. In the field of information systems, many researchers recommend to incorporate individual characteristics into a research model either as control variables or as independent variables to study the cognitive, affective, and/or behavioral reactions of individuals to technology [22, 42, 68]. Personal characteristics include salient (or observable) attributes such as age and gender, and subtle (or unobservable) attributes such as education and personality [49]. In e-commerce, the type of personality that receives most research attention is one s propensity to trust, or the tendency to believe or not to believe in others and so trust them [23]. Characteristics-based trust antecedents are most relevant in the initial stage of trust building, when individuals have no direct experience with the target e-vendor [23, 44]. 2) Knowledge-based Trust Antecedents: Knowledge-based trust antecedents include knowledge, experience, and self-efficacy with online transactions. These factors can find their theoretical roots in human capital model [9] of marketing and social cognitive theory [6] of sociology. Human capital model treats consumer knowledge as human capital, which affects search activities and consumption prices [9]. Human capital can be acquired through investments in formal and informal education, training, and learning by doing [57]. In contrast, social cognitive theory centers on the concept of self-efficacy as the main mechanism to regulate one s behavior. Selfefficacy, defined as beliefs about one's ability to perform a specific behavior, is induced from psychological procedures of deliberating information from various sources including direct and vicarious experiences [6, 7]. Both theories contend that knowledge and experience with target behaviors will help individuals to behave efficiently and decisively in similar contexts. In e-commerce, the more knowledge and experience one has accumulated with a website, the more trust one is likely to form in that website [23]. 3) Deterrence-based Trust Antecedents: Deterrence-based trust is based on the assumptions that individuals are rational and will not engage in opportunistic behavior whose consequences of being untrustworthy are possibly severe [60]. The assumptions are in line with the transaction cost theory, which presumes that economic actors behave with bounded rationality and self-interest [61, 74]. As such, individuals can calculate the consequences of another party s cheating or cooperating in the relationship, and form beliefs whether the party can be trusted (or, unlikely to perform untrustworthy activities). In the research of e-commerce, main types of deterrence-based trust include calculative trust and legal framework. Calculative trust is derived from classic economic theories with a utilitarian tradition [26]. Trust is viewed as a result of rational calculation of individual gains [73]. In the context of e-commerce, a consumer can trust an e-vendor if he/she believes that the e-vendor has more to lose than to gain by cheating or has nothing to gain by breaking customer trust [23, p. 64]. Legal framework also falls in the category of deterrence-based trust antecedents. Legal framework refers to the perceived regulations and law enforcements in online transactions [14]. Consumers tend to develop trust in e-vendors if they believe the existing laws and regulations are adequate to protect their interests. E-vendors are unlikely to perform untrustworthy activities in the fear of severe punishments. 4) Social Influence-based Trust Antecedents: The influence of social factors on trust can find its theoretical ground in the social theory of human behavior. Contrast to economists view of human behavior as resulting from rationality and self-interest, sociologists emphasizes on the role of social relations [26], and frequently call attention to the intensity with which men desire and strive for the good opinion of their immediate associates in a variety of situations [75, p. 189]. People s purposive actions are constrained by concrete, ongoing systems of social relations including social networks, culture, politics and religion [27]. In the research of e-commerce, social norms [39, 68], ratings from fellow consumers [79], and references from friends (i.e., word-of-mouth) [70] have been studied as important social factors that affecting the formation and development of trust. 5) Technological Attributes-based Trust Antecedents: Viewing an e-vendor s website as a special information system, online transaction behaviors can be partially explained by the technological attributes of the website [23]. Technological attributes derived from the research of technology acceptance model (TAM) have been widely incorporated into trust model as antecedents. These technical factors include perceived ease of use and perceived usefulness [15, 16], perceived enjoyment [30, 36], website quality [38, 44], and privacy and security protections [70]. All these factors reflect the technical wellness of an e-vendor in terms of supporting transactions. 6) Vendor-image-based Trust Antecedents: Derived from consumer behavior theories, vendor-image-based trust antecedents include reputation, company size, and product choices of a target e-vendor. These factors are able to convey to consumers an image of the e-vendor, helping them develop brand awareness and make favorable decisions toward the e-vendor by employing simple heuristic choices [33]. These vendor-image factors are particularly important for first-time visitors, who have no previous experience with the target website but rely heavily on website cues and company reputation to form their initial trust beliefs [44]

3 7) Institution-based Trust Antecedents: In the research of e-commerce, institution-based trust refers to a buyer s perception that third-party guarantees are in place to facilitate online transactions [54]. Certain design of IT infrastructure and artifacts, such as strategic alliance, third-party assurance and certification (including credit card guarantees), can effectively boost institution-based trust and create conditions that will facilitate transaction success [1, 24]. The mechanism can be explained by the signaling theory that assumes information asymmetry in transactions [65]. According to the signaling theory, an under-informed party may employ signals for information of the other party to resolve the asymmetry. In job market for example, employers often use education degree as a credible signal for an applicant s ability to learn [64]. In e-commerce, guarantees from a third party may serve as a strong signal of trustworthiness if the third party is perceived as an expert that specializes in certifying secured online transactions. As Aiken and Boush [1] pointed out in their study of Internet signals: in the context of evaluating an unknown Internet-based firm, consumers are likely to weigh such third-party information heavily [1, p. 311]. Situational normality (similar concepts include compatibility in [48, 54], and perceived fairness in [55]), defined as the assessment that the transaction will be a success based on how normal or customary the situation appears to be [4], also falls into the category of institution-based trust antecedents [23]. Situational normality helps assure people that everything in a standard setting is as it ought to be [47] and strange behaviors (e.g., cheating) are unlikely to happen. C. The Effects of Trust Consequents In contrast with the complex nature of the formation and development of trust, trust consequents that have been studied in e-commerce are comparatively cohesive encouraging online consuming activities by affecting the attitudes, intentions, and eventually the actual behaviors of online transaction [23]. Trust effects on these consequents can find strong support in behavior research about behavior formation (e.g., the theory of reasoned action [2]) and psychology research about behavior regulation and change (e.g., the social cognitive theory [6]). In addition, TAM also provides theoretical support that favorable beliefs, i.e., perceived ease of use and perceived usefulness of an investigated technology [15, 16], help to form positive attitudes including trust [23] toward the technology, and ultimately lead to desired behaviors. D. An Integrated Model of Trust The discussion of trust antecedents and trust consequents is summarized in Fig. 1. The model integrates different theoretical perspectives regarding the formation and development of trust in e-vendors without probing possible interwoven relationships among the suggested factors. The model also implies that trust plays an important role in e-commerce by mediating the effects of various factors on individual behaviors in online transactions. Fig. 1 An Integrated Model of Trust in E-commerce III. RESEARCH METHOD META-ANALYSIS This study attempts to systematically review what we have empirically examined in the literature of trust in e-commerce. Meta-analysis is employed for its unique strength of viewing the whole picture in a research context by combining and analyzing the quantitative results of many empirical studies [25]. Compared with other literature review methods such as narrative review, meta-analysis is less judgmental and subjective with a focus on data and the use of rigorous statistical procedures [43]. A. Selection of Relevant Studies To systematically sample relevant studies, two popular business research databases ABI/INFORMS and Social Science Citation Index (SSCI) were searched using the keywords trust and e-commerce or online transactions. In addition, the

4 AIS conference database was also searched to alleviate the concern of publication bias [43]. There are other sources which should be included to ensure the comprehensiveness of the search, e.g.: EBSCO, Science Direct, Emerald, ACM, Springer, etc. The fact that these databases overlap with each other especially for the most influential research outlets alleviates the concern that some significant findings were ignored from the search. The exhaustive search resulted in 274 articles that were published since Among the 274 articles, 204 articles were eliminated because they (a) were not empirical studies, (b) did not study e-commerce, (c) did not measure trust, or (d) did not report the effects (as explained below) that can be used for the meta-analysis. The resulting 70 studies provided 262 effect sizes for this meta-analysis. The earliest empirical study of trust in e-commerce was found to be Gefen s [21] work published in 2000 on the importance of familiarity to the formation of trust; the latest one included in the sample is Weisberg and colleagues [72] work published in January 2011 examining the mediating role of e-vendors social presence on one s development of trust. The scope of the meta-analysis is the most exhaustive one to date. Interested readers can contact the author for a complete list of the sampled studies and their coded effect sizes. The systematical search may raise a concern that the focal construct, trust, is measured differently across the sampled studies. As aforementioned, Mayer et al. s [45] conceptualization of trust dominates the research of trust in e-commerce. However, other conceptualizations also exist in the literature. For example, Gefen et al. [23] have discussed that some researchers view trust as a general belief that the other party can be trusted, or a sentiment reflected in the feelings of confidence and security with the other party. Accordingly, various instruments have been employed for measuring trust, ranging from single-item rating (e.g., [79]) or index (e.g., [28]), to multi-item instruments (e.g., Nicolaou and McKnight [51] developed 9 items to measure trust believes). Several researchers measured trust as a second-order factor (e.g., [19, 36, 47, 63, 71]) that is derived from the original conceptualization in Mayer et al. [45]. In experimental studies, trust can be manipulated by the design of task environments (e.g., [53]). Nonetheless, all the sampled studies view trust as the willingness to take risks or being vulnerable to the actions of another party; the difference focuses on the expectation of possible actions from the other party (i.e., the trustee) [39]. In this sense, a systematic integration of the findings will provide implications that can be generalized to the research of trust. In fact, one criticism of meta-analysis is that it may compare apples and oranges by aggregating results derived from studies with incommensurable research goals, measures, and procedures. This problem is not of great significance here, since the attempt is to obtain results that are generalizable to a fairly broad domain [43] of trust research. B. The Treatment of Effect Sizes Effect sizes used in this meta-analysis were Pearson s correlation coefficients. Most studies reported correlation coefficients between trust and antecedents/consequents. If such correlation coefficients were not available, other statistics (e.g., descriptive statistics, p-values, T-tests, F-tests, and d-tests) were examined for possible conversion to correlation coefficients by following the procedures described by Rosenthal and DiMatteo [58]. Some studies reported the detailed correlation matrix at the instrument item level [21]. The matrix was extracted and LISREL was used to calculate the relationship of interest. If a study had more than one effect size regarding a particular relationship described with the previously-discussed classification of variables, the effect sizes were treated by following the decision rules suggested by King and He [41]: multiple effect sizes are averaged if they are homogeneous in both magnitudes and research contexts; otherwise, they will be treated separately. C. Meta-analytic Calculations and Results This meta-analysis collected 262 effect sizes from 70 studies. The data set was first screened for possible outliers. An effect size was judged as an outlier if it did not represent the population of interest with an (1) extraordinarily large effect size, or (2) extraordinarily large sample size. Following the rules, seven effect sizes were judged as outliers and were excluded from further analysis. Although individual case outliers are not a substantial source of residual variability for the integrated effect size [52], removing outliers is believed to improve the reliability of meta-analytic estimations [8]. After removing outliers, Hedges-Olkin s [31] random effects method was used as the calculation procedure. Hedges-Olkin is one of three well-developed meta-analytic procedures that have been widely applied in the social and behavioral sciences [17, 18]. Basically, effect sizes are first transformed into Fisher s z; weights of individual studies are calculated based on sample sizes (using a chi-square analysis of sampling errors); a homogeneity test (Q test) is performed to justify the use of an effect-size-integration model (in this case, the random model); then, a weighted average is calculated from transformed effect sizes, the associating standard deviation, Z-statistic, and 95% confidence intervals. Finally, these statistics are transformed back to their standard forms for interpretation. Results of the meta-analysis are presented in Tables I-VII for trust antecedents, and Table VIII for trust consequents

5 TABLE I PERSONAL CHARACTERISTICS-BASED ANTECEDENTS OF TRUST Gender Age Education Trust Propensity Income # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) < % Low (r) % High (r) At the first glance, all homogeneity tests report insignificance with large p-values (>0.20). The homogeneity test is a test of the null hypothesis that the interaction error term (between the sample error and the study error) is zero [51]. Insignificance of testing results indicates that heterogeneity among sampled studies cannot be ruled out after controlling sample errors, therefore validating the use of a random effects analytic base [30] for the current study. Despite the call for incorporating personal characteristics into IS research [40, 69, 112], not many studies have examined the effects of demographics and personalities in their research of trust in e-commerce. The extensive search in the literature has identified 25 studies in this particular area. Much emphasis has been placed on trust propensity (17 out of 25 sampled studies); other demographic factors, including gender, age, education, and income, have received le research attention. The direct effects of gender, age, and education on trust are found to be frivolous (r close to 0), while trust propensity is found to exert a notable, although small in magnitude, influence on trust (r=0.222, p<0.001). Interestingly, income level is also found to significantly correlate with one s trust in e-commerce; high-income individuals are more likely to trust e-vendors and purchase online than low-income individuals are [97]. However, the lack of repeated empirical tests in various research contexts excludes the factor from being meta-analyzed. TABLE II KNOWLEDGE-BASED ANTECEDENTS OF TRUST Familiarity with E-vendor Self-efficacy Internet Experience # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) < % Low (r) % High (r) Knowledge-based antecedents of trust include familiarity with the e-vendor, self-efficacy, and Internet experience. The former two factors are found to be strong antecedents of trust, with notable effect sizes (r=0.329 and respectively) and strong statistical significance (p<0.001 and p=0.003 respectively). Interestingly, Internet experience is found to have no significant effects on the formation and development of trust in e-commerce (r=0.069, p=0.425). This finding suggests that trust in e-commerce is mostly derived from the special knowledge of target e-vendors (i.e., familiarity with e-vendor) and confidence in conducting online transactions (i.e., self-efficacy), and is little affected by the general knowledge of Internet. In other words, knowing Internet is an insufficient condition for one to embrace e-commerce if the person does not have the special knowledge of conducting transactions with certain e-vendors. TABLE III DETERRENCE-BASED ANTECEDENTS OF TRUST TABLE IV SOCIAL-BASED ANTECEDENTS OF TRUST Calculative Trust Legal Framework # of sampled studies 5 5 Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) <0.001 < % Low (r) % High (r) Reference/ Word-of-mouth Negative Rating Social Norms # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) < % Low (r) % High (r) In the literature, two factors calculative trust and legal framework have been examined as deterrence-based antecedents of trust. The two factors present strong influence on trust (r=0.453 for calculative trust, and r=0.358 for legal framework; both with p<0.001). The results suggest that people are likely to trust a web-site if they are convinced that the web-site is held accountable for any misbehavior either by economic means or by law enforcements. Social-based antecedents of trust include reference/word-of-mouth from others, negative rating, and social norms. Of these

6 antecedents, negative rating presents a strong negative effect on trust (r=-0.46, p<0.001), highlighting the detrimental effects of disapproving information to a web-site. Contradictory to common understanding, references or word-of-mouth from others demonstrate a marginal effect on trust (r=0.264, p=0.087). As for social norms, no study has been found to address the particular factor. TABLE V TECHNOLOGICAL ATTRIBUTES-BASED ANTECEDENTS Perceived Ease of Use Perceived Usefulness Website Quality Perceived Enjoyment Feedback Mechanisms Privacy & Security # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) <0.001 <0.001 <0.001 < < % Low (r) % High (r) Studying the technological attributes of a target e-vendor s website is found to be popular in the research of trust in e- commerce. The factors that have been examined in the literature include perceived ease of use, perceived usefulness, website quality, perceived enjoyment, feedback mechanisms, privacy and security. Overall, these technological attributes are found to be very influential for one to develop trust beliefs toward the target website. The statistical significances of these technological attributes are impressively high(p<0.001for most factors, except p=0.007 for feedback mechanisms); in addition, the magnitudes of their effect sizes range from to 0.538, and can be judged as moderately large by common heuristics used in the social and behavioral sciences [13]. TABLE VI VENDOR-IMAGE-BASED ANTECEDENTS OF TRUST TABLE VII INSTITUTION-BASED ANTECEDENTS Reputation Company Product Size Choices # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) < % Low (r) % High (r) Assurance /Certification Situational Normality # of sampled studies 9 6 Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) <0.001 < % Low (r) % High (r) The comprehensive search of empirical studies in the literature of trust in e-commerce shows that IS researchers have placed limited attention on the vendor-image-based antecedents such as reputation, size, and product choices. The latter two factors have each received only one empirical test in the literature. However, the three factors demonstrate remarkable effect sizes, with reputation having r=0.449, company size having r=0.44, and product choices having r=0.21. More research attention needs to be placed on this group of factors for studying trust in e-commerce. In a conference panel discussion, Gefen and colleagues [24] highlighted the importance of studying institutional trust in IS research. Since then, institution-based antecedents of trust have received increased attention among IS researchers. However, as demonstrated in Table VII, limited attempts have been conducted in the context of e-commerce. Similar to that of technology-based antecedents, the two institution-based antecedents present strong influences on the formation and development of trust in e-commerce, with notably large effect sizes (0.405 for third-party assurance and certification, and for situational normality) and strong significance (p<0.001). More research attention needs to be placed on this group of factors in the future research of trust in e-commerce. TABLE VIII CONSEQUENTS OF TRUST Affecting Transaction Behaviors Reshaping Customers Perceptions Attitude Behavioral User Price Recommendation Perceived Perceived Perceived Behavior Intention Loyalty Premiums (Word-of-Mouth) Risk Usefulness Ease of Use # of sampled studies Total sample size Homogeneity Test (Q) p (heterogeneity) Avg (r) Z p (effect size) <0.001 < < <0.001 < % Low (r) % High (r) The importance of studying trust in e-commerce centers on the arguments that trust facilitates online consuming activities

7 by affecting the attitudes, intentions, and eventually the actual behaviors of online transaction [23]. Thus, the constructs of attitude, behavioral intention, and/or actual behavior are widely accepted as key consequents of trust in the e-commerce literature of trust research. In the meta-analysis, 47 out of 52 studies that have tested trust effects in e-commerce select one or more of the three constructs as the dependent variables in their empirical investigations. Some researchers have extended their investigations to other trust-triggered attitudes and behaviors such as user loyalty, the acceptance of price premiums, and making recommendations to other Internet shoppers. Overall, trust exerts large effects on attitudinal and intentional factors (absolute effect sizes ranging from to 0.527), and guides one s behavior in making recommendations to other Internet shoppers (Avg(r) = 0.621); the effects on actual transaction behavior is found to be moderately small in magnitude (Avg(r) = 0.189) but reasonably strong in significance (p=0.038). These findings lend support to the importance of studying trust in the research of e-commerce. With limited attempts (8 out of the 52 sampled studies that have tested trust effects in e-commerce), the literature has tested the extent to which trust reshapes consumers perceptions of target e-vendors. The effects of trust on perceived risks and perceived usefulness are found to be large. In contrast, perceived ease of use of a website is found to be marginally affected by the development of trust (Avg(r) = 0.231, p=0.15). IV. DISCUSSIONS This study intends to synthesize the empirical findings of previous trust studies in the context of e-commerce. 70 relevant studies were identified from the literature, and 262 effect sizes were collected and analyzed. The results demonstrate the importance of trust as a key intervening mechanism inshaping one s behaviors in the context of e-commerce. As many as 24 trust antecedents were analyzed. The broad range of antecedents suggests that the formation of trust is influenced by affective, cognitive, technological, and social forces beyond the boundary of any single theory. These trust antecedents are reported in Table IX sorted by the size of each integrated effect (i.e., Avg(r)). Following Cohen s [19] guideline, the effects are grouped in four categories of large (Avg(r) close to 0.5), medium (Avg(r) close to 0.3), small (Avg(r) close to 0.1), and little effects (Avg(r) close to 0). TABLE IX ANTECEDENTS OF TRUST SORTED BY EFFECT SIZES Effect Size Levels Antecedents # of Sampled Studies Total Sample Size Average Effects (r) p-value 1. Perceived Usefulness < Situational Normality < Perceived Ease of Use < Website Quality < Negative Rating <0.001 Large 6. Privacy & Security <0.001 Effects 7. Calculative Trust < Reputation < Company Size Assurance/Certification < Perceived Enjoyment < Feedback Mechanisms Legal Framework <0.001 Medium Effects 14. Familiarity with e-vendor < Self-Efficacy Reference/Word-of-mouth Small Effects LittleEffects 17. Trust Propensity < Product Choices Income Internet Experience Education Gender Age Not Studied 24. Social Norms 0 0 Among the investigated factors, deterrence-based, social and institution-based, and technological attributes-based antecedents are found to be the most influential forces on the formation and development of trust. The magnitude of their effects falls in the range of , which can be judged as medium to large by common heuristics used in the social and behavioral sciences [13]. This finding suggests that trust in e-commerce is largely in line with technology usage [20, 23], cognitive and rational transaction behavior [57, 60], and social and institutional influences [26, 54], probably because e- commerce is in nature a technology-driven and transaction-driven environment. Some factors have received limited attention in the literature. The exhaustive search of trust studies in e-commerce has not located any test for the effect of social norms, a widely acknowledged factor for shaping individual s behavior in technology

8 adoption and usage [69]. Respectively, the search has located only one test for the effects of company size, product choices, and income, two tests for the effects of education and gender, and three tests for the effect of age. As results from few studies cannot serve as the basis for generalizable conclusions, future examination of these factors should be encouraged. The accumulation of empirical evidence is a necessary process for reaching conclusive results and an essential condition for the field to progress [35]. Overall, personal characteristics other than trust propensity have received least attention in the research of trust in e- commerce. In fact, personal characteristics have been reported in many studies mainly for the purpose of demographic analysis of a target research sample. This is most obvious in culture-oriented studies that compare the attitudes and behaviors of consumers across different countries (e.g., [62, 76]). This observation is surprising. Examining the effects of personal characteristics is common in many areas of IS research including computer self-efficacy (e.g., [5, 12, 37, 40]) and technology acceptance (e.g., [50, 66, 68]). Especially gender, a salient demographic factor, has been widely recognized to have a profound influence on one s reactions to technology [50]. Future research needs to place more attention on the influence of personal characteristics on trust in e-commerce. In addition, the small direct effects of personal characteristics on trust need further investigation. The effects of personal characteristics observed in the literature are typically small or negligible. Even the effect of trust propensity, which is widely acknowledged as a strong trust antecedent evidenced by the large number of empirical tests, is small in magnitude (Avg (r) =0.222). The influence of personal characteristics on trust is based on the assumption that people with similar backgrounds and experiences will behave in similar ways. Such a contention suggests that personal characteristics exert a moderating role in the development of trust. However, there is rare endeavor in the literature to examine personal characteristics as moderators rather than determinantors of trust in e-commerce. One exception is [56], in which Qureshi and colleagues found that gender and age indirectly influence one s trust in e-vendors by affecting his/her perception of vendors website and service quality. In another study, Zhang and colleagues [78] examined the moderating role of gender in their trust model. Gender was found to moderate the effects of two trust antecedents of perceived ease of navigation and perceived web presentation quality; however, the direct relationship between gender and trust, or the correlation, was not reported. Thus, studying the mechanisms through which personal characteristics affect individual trust beliefs and behaviors will be a promising area for future research. Another interesting finding of the meta-analysis is the controversial effect of Internet experience. Of the seven sampled studies that had empirically tested the influence of Internet experience on trust, three studies reported significant positive effects, two studies reported significant negative effects, and two studies reported insignificant effects. The aggregated effect of Internet experience on trust is trivial (Avg (r) = 0.069, p=0.425). In another study, Pizzutti and Fernandes [55] found that prior experience with e-commerce significantly affected one s trust in on-line seller (r=0.18, p<0.01). The study was judged as an outlier because of its uncommonly large sample size (N=3206, while the average sample size of other studies in the group is 454). If Pizzutti and Fernandes [55] finding was included in the analysis, the integrated effect size would be slightly increased (Avg (r) = 0.087, p=0.186); still, no significance could be concluded. This finding is contradictory to the common wisdom that a better knowledge of e-commerce will induce one to trust e-vendors and conduct more transactions over the Internet. Given the conflicting findings in the literature, it still remains an unsolved question whether consumers will be less skeptical with increased experience of e-commerce. Research addressing this issue will provide valuable insights on the cognitive process of trust development. As for the consequents of trust, the literature is quite coherent on what trust may affect in the context of e-commerce. Most empirical studies select consumers attitude, behavioral intention, and/or actual behavior as the dependent variables. Some studies have tested other trust-triggered attitudes and behaviors such as user loyalty, the acceptance of price premiums, and making recommendations to other Internet shoppers. As demonstrated in Table VIII, the overall effects of trust are strong in both magnitude (most integrated effect sizes fall in the range of ) and statistical significance, lending support to the importance of studying trust in the research of e-commerce. Some researchers have studied the extent to which trust re-shapes one s perceptions toward a target website such as perceived risk, perceived usefulness, and perceived ease of use (Table VIII). These studies reinforce the reciprocal relationship between behavior and attitudes. As Bem [10] pointed out, the behavioral experiences, if they are mindful and reflect the person s willingness, define and change attitudes, and direct the performance of future behaviors. Thus, the antecedents-trustbehavior relationship can therefore be viewed as a comprehensive interactive system in which the antecedents, trust, and behaviors exert reciprocal influences on one another. The investigation of the interactive system especially on the reciprocal relationships between trust and trust antecedents will strengthen the theoretical foundations of trust research in e-commerce. Some limitations of the study should be acknowledged. One limitation is about the coverage of sampled papers. The exhaustive search does not cover all research databases (e.g., Business Source Premier (an EBSCO Host database)) and may have failed to found some articles online or through the employed university research library system. However, given the overall large number of sampled studies, adding a few additional studies would be unlikely to alter the conclusions. Another limitation is about the treatment of effect sizes. This study used reported effect sizes without adjustments on their measurement errors, such as perception-perception bias, single-method bias, and attenuation effect caused by imperfect construct reliabilities. Techniques of estimating and reducing these measurement errors are available, but require additional

9 examination of each sampled study of its detailed research setting and methods, and probably over-modified statistical procedures, which are beyond the scope of the current study. Both the number of the sampled studies and the total cumulated sample size are satisfactorily large, granting credibility to these results. V. CONCLUSIONS This meta-analysis of 70 empirical trust studies provides powerful large-sample evidence that: (1) trust plays a critical role in shaping one s attitudes and behaviors in the conduction of online transactions; (2) deterrence-based, social and institutionbased, and technological attributes-based antecedents are found to be most influential in the formation of trust; (3) personal characteristics are in general understudied; some salient demographic factors such as gender and age are likely to moderate rather than to directly affect the formation of trust among online consumers. REFERENCES [1] Aiken, K.D., Boush, D.M. Trustmarks, Objective-source ratings, and implied investments in advertising: Investigating online trust and the context-specific nature of Internet signals, Academy of Marketing Science, 34(3), pp , [2] Ajzen, I. Attitudes, traits, and actions: Dispositional predictions of behavior in personality and social psychology, Advances in Experimental Social Psychology, 20, pp. 1-63, [3] Ba, S. and Pavlou, P.A. Evidence of the effect of trust building technology in electronic markets: Price premiums and buyer behaviour, MIS Quarterly, 26(3), pp , [4] Baier, A. Trust and antitrust, Ethics, 96(2), pp , [5] Ballou, D.J., and Huguenard, B.R. The impact of students' perceived computer experience on behavior and perform in an introductory information systems course, Journal of Information Systems Education, 19(1), pp , [6] Bandura, A. Self-efficacy: Toward a unifying theory of behavioral change, Psychological Review, 84(2), pp , [7] Bandura, A. Self-efficacy mechanism in human agency, American Psychologist, 37, 2 (1982), [8] Beal, D.J., Corey, D.M., and Dunlap, W.P. On the bias of Huffcutt and Arthur's (1995) procedure for identifying outliers in the metaanalysis of correlations, Journal of Applied Psychology, 87(3), pp , [9] Becker, G., Human Capital, Chicago: University of Chicago Press, [10] Bem, D.J. Self-perception theory, Advances in Experimental Social Psychology, 6, pp. 1-63, [11] Bhattacherjee, A. Individual trust in online firms: scale development and initial test, Journal of Management Information Systems, 19(1), pp , [12] Bozionelos, N. Psychology of computer use: XXXIX. Prevalence of computer anxiety in British managers and professionals, Psychological Reports, 78, pp , [13] Cohen, J. A power primer, Psychological Bulletin, 112(1), pp , [14] Connolly, R., and Bannister, F. Consumer trust in Internet shopping in Ireland: Towards the development of a more effective trust measurement instrument, Journal of Information Technology, 22(2), pp , [15] Davis, F.D. Perceived usefulness, perceived ease of use, and user acceptance of information technology, MIS Quarterly, 13(3), pp , [16] Davis, F.D., Bagozzi, R.P., and Warshaw, P.R. User acceptance of computer technology: A comparison of two theoretical models, Management Science, 35(8), pp , [17] Field, P. Meta-analysis of correlation coefficients: A Monte Carlo comparison of fixed- and random-effects methods, Psychological Methods, 6(2), pp , [18] Field, P. The problem in using fixed-effects models of meta-analysis on real-world data, Understanding Statistics, 2(2), pp , [19] Flavian, C., Guinaliu, M. and Gurrea, R. The role played by perceived usability, satisfaction and consumer trust on website loyalty, Information & Management, 43(1), pp. 1-14, [20] Friedman, B., Khan, P.H. Jr., and Howe, D.C. Trust online, Communications of the ACM, 43(12), pp , [21] Gefen, D. E-commerce: The role of familiarity and trust, Omega, 28(6), pp , [22] Gefen, D., and Straub, D.W. Gender differences in the perception and use of An extension to the technology acceptance model, MIS Quarterly, 21(4), pp , [23] Gefen, D., Karahanna, E., and Straub, D.W. Trust and TAM in online shopping: An integrated model, MIS Quarterly, 27(1), pp , [24] Gefen, D., Pavlou, P., Benbasat, I., McKnight, H., Stewart, K., and Straub, D.W. ICIS panel summary: Should institutional trust matter in information systems research? Communications of the Association for Information Systems, 17(9), pp , [25] Glass, V. Primary, secondary, and meta-analysis of research, Review of Research in Education, 5(10), pp , [26] Granovetter, M. Economic action and social structure: The problem of embeddedness, The American Journal of Sociology, 91(3), pp , [27] Granovetter, M. The impact of social structure on economic outcomes, Journal of Economic Perspectives, 9(1), pp , [28] Grimsley, M. and Meehan, A. E-government information systems: Evaluation-led design for public value and client trust, European Journal of Information Systems, 16(2), pp ,

10 [29] Guo, C., Shim, J.P., and Otondo, R. Social network services in China: An integrated model of centrality, trust, and technology acceptance, Journal of Global Information Technology Management, 23(2), pp , [30] Hampton-Sosa, W., and Koufaris, M. The effect of web site perceptions on initial trust in the owner company, International Journal of Electronic Commerce, 10(1), pp , [31] Hedges, L.V., and Olkin, I. Statistical Methods for Meta-Analysis. San Diego: Academic Press, [32] Hoffman, L., Novak, T.P., and Peralta, M. Building consumer trust online, Communications of the ACM, 42(4), pp , [33] Hoyer, W.D., and Brown, S.P. Effects of brand awareness on choice for a common, repeat-purchase product, Journal of Consumer Research, 17(2), pp , [34] Hsiao, K.L., Lin, J.C.C., Wang, X.Y., Lu, H.P., Yu, H. Antecedents and consequences of trust in online product recommendation: An empirical study in social shopping, Online Information Review, 34(6), pp , [35] Hunter, J.E., Schmidt, F.L., and Jackson, G.B. Meta-Analysis: Cumulating Research Findings Across Studies. Beverly Hills: Sage, [36] Hwang, Y., and Kim, D.J. Customer self-service systems: The effects of perceived Web quality with service contents on enjoyment, anxiety, and e-trust, Decision Support Systems, 43(3), pp , [37] Igbaria, M., and Chakrabarti, A. Computer anxiety and attitudes towards microcomputer use, Behavior and Information Technology, 9(3), pp , [38] Jones, K., and Leonard, L.N.K. Trust in consumer-to-consumer electronic commerce, Information & Management, 45(2), pp , [39] Karahanna, E., and Straub, D.W. The psychological origins of perceived usefulness and ease-of-use, Information & Management, 35(4), pp , [40] King, J., Bond, T., and Blandford, S. An investigation of computer anxiety by gender and grade, Computers in Human Behavior, 18(1), pp , [41] King, W.R., and He, J. A meta-analysis of the technology acceptance model, Information & Management, 43(7), pp , [42] King, W.R., and He, J. External validity in IS survey Research, Communications of the Association for Information Systems, 16, pp , [43] King, W.R., and He, J. Understanding the role and methods of meta-analysis in IS research, Communications of the Association for Information Systems, 16, pp , [44] Koufaris, M., and Hampton-Sosa, W. The development of initial trust in an online company by new customers, Information & Management, 41(3), pp , [45] Mayer, R.C., Davis, J.H., and Schoorman, F.D. An integrative model of organizational trust, Academy of Management Review, 20(3), pp , [46] McCloskey, W. The importance of ease of use, usefulness, and trust to online consumers: An examination of the technology acceptance model with older consumers, Journal of Organizational and End User Computing, 18(3), pp , [47] McKnight, D.H., Choudhury, V., and Kasmar, C. Developing and validating trust measures for e-commerce: An integrative typology, Information Systems Research, 13(3), pp , [48] McKnight, D.H., Cummings, L.L., and Chervany, N.L. Initial trust formation in new organizational relationships, The Academy of Management Review, 23(3), pp , [49] Milliken, J., and Martins, L.L. Searching for common threads: Understanding the multiple effects of diversity in organizational groups, Academy of Management Review, 21(2), pp , [50] Morris, M.G., Venkatesh, V., and Ackerman, P.L. Gender and age differences in employee decisions about new technology: An extension to the theory of planned behaviour, IEEE Transactions on Engineering Management, 52(1), pp , [51] Nicolaou, I., and McKnight, D.H. Perceived information quality in data exchanges: Effects on risk, trust, and intention to use, Information Systems Research, 17(4), pp , [52] Orr, J.M., Sackett, P.R., and DuBois, C.Z. Outlier detection and treatment in I/O psychology: A survey of researcher beliefs and an empirical illustration, Personnel Psychology, 44, pp , [53] Parasuraman, R., and Miller, C.A. Trust and etiquette in high-criticality automated systems, Communications of the ACM, 47(4), pp , [54] Pavlou, P.A., and Gefen, D. Building effective online marketplaces with institution-based trust, Information Systems Research, 15(1), pp , [55] Pizzutti, C, and Fernandes, D. Effects of recovery efforts on consumer trust and loyalty in E-tail: A contingency model, International Journal of Electronic Commerce, 14(4), pp , [56] Qureshi, I., Fang, Y., Ramsey, E., McCole, P., Ibbotson, P., and Compeau, D. Understanding online customer repurchasing intention and the mediating role of trust - an empirical investigation in two developed countries, European Journal of Information Systems, 18(3), pp , [57] Ratchford, B.T. The economics of consumer knowledge, Journal of Consumer Research, 27(4), pp , [58] Rosenthal, R., and DiMatteo, M.R. Meta-analysis: Recent developments in quantitative methods for literature reviews, Annual Review of Psychology, 52(1), pp , [59] Rousseau, D.M., Sitkin, S.B., Burt, R.S., and Camerer, C.C. Not so different after all: A cross-discipline view of trust, Academy of Management Review, 23(3), pp , [60] Shapiro, D.L., Sheppard, B.H., and Cheraskin, L. Business on a handshake, Negotiation Journal, 8(4), pp ,