Electronic Commerce Research and Applications 11 (2012) 335–345

Contents lists available at SciVerse ScienceDirect

Electronic Commerce Research and Applications

journal homepage: www.elsevier .com/locate /ecra

How to improve consumer attitudes toward using credit cards online: Anexperimental study

Wesley Shu, Chiang Yu Cheng ⇑National Central University, No. 300, Jhongda Rd., Jhongli City, Taoyuan Country 32001, Taiwan

a r t i c l e i n f o

Article history:Received 31 March 2011Received in revised form 22 January 2012Accepted 22 January 2012Available online 2 February 2012

Keywords:Attitude changeBehavioral researchCredit cardsConsumer behaviorOnline shoppingMessage proximityMessage sourcePersuasion theorySocial justice theoryTransaction assurance

1567-4223/$ - see front matter � 2012 Elsevier B.V. Adoi:10.1016/j.elerap.2012.01.003

⇑ Corresponding author. Tel.: +886 953676721.E-mail address: taican.ccy@gmail.com (C.Y. Cheng

a b s t r a c t

Most companies involved in e-commerce use trust-promoting messages to persuade consumers that thewebsites they visit are safe. Although these messages are common, many consumers still hesitate to use acredit card online. To understand why, we applied social judgment theory and Cialdini’s model of persuasionin an experiment exploring the extent to which trust-promoting messages can be used to change con-sumers’ attitudes toward online credit card use. By manipulating message proximity and message source,we found that authority-plus-contrast-plus-scarcity was the most persuasive combination, and there wasno significant difference between international and domestic trustmarks. Consumers with positive creditcard attitudes were included as a control condition. The results show that the authority-plus-contrast-plus-scarcity combination was essential for consumers holding the same positive attitude, because therewas no significant attitude change in this condition. Our findings help shopping websites analyze theirexisting trust-promoting messages and improve them by adding persuasive elements.

� 2012 Elsevier B.V. All rights reserved.

1. Introduction

The topic of trusting attitudes has received wide attention inthe context of online shopping (McKnight and Chervany 2001,Heijden et al. 2003). These attitudes are defined as a consumer’snegative or positive feelings about making a purchase (Crespoand Bosque 2010). The purpose of the study reported in this paperis to analyze trusting attitude in using credit cards online. One wayto improve this attitude is to provide trust-promoting messages onthe website’s check-out page (i.e., self-proclaimed assurances). Anexample of a trust-promoting message found on Coast to CoastTickets.com’s website (www.coasttocoasttickets.com) is: ‘‘We pro-vide a 100% safe transaction environment for our consumers.’’

Because consumers cannot fully monitor the trustworthiness ofshopping websites, trust-promoting messages generated by a thirdparty (trustmarks) serve the role of gatekeeper (Benassi 1999). Atrustmark, as its name implies, is a mark, logo, seal, or symbolaimed at building consumers’ confidence in the security and pri-vacy of online information (Aiken et al. 2003). One of the leadingproviders of such information protection is VeriSign. Used byaround 93% of Fortune 500 companies, VeriSign is the most visibletrustmark in the industry, recognized by 79% of online consumers

ll rights reserved.

).

in the US (TNS 2008). Even though trustmarks are widely displayedon shopping websites, 75% of online consumers still hesitate togive out their credit card numbers. In another survey (Boone andKurtz 2010), it was found that only 28.3% of consumers used acredit card online, even though 83% of Internet stores providedboth first-party and third-party assurances on their websites(MIC 2010). These results indicate that consumers’ negative atti-tudes toward using credit cards online is still one of the majorobstacles to online shopping.

To improve consumers’ attitudes about using a credit cardonline, we first need to understand how the gap can be narrowedbetween consumers’ pre-existing attitudes and the message’sproclamation. One way to accomplish this is to learn how to pres-ent the messages so consumers will gradually come to acceptthem. The next issue is what factors (e.g., message content, mes-sage proximity, message source) we should pay attention to inmodifying the message. We first refer to social judgment theory,according to which recipients’ responses to persuasive messagesare determined by their pre-existing attitudes and the attitudepromoted by the message (Sherif and Hovland 1980). We call thecloseness of these two attitudes message proximity. Second, wedraw upon the three component principles (authority, contrast,and scarcity) from Cialdini’s (1993) research on persuasion. Therole of these components is to explain the impact of message prox-imity on consumer attitude change.

336 W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345

The goal of all this is to find the best way to change consumerattitudes toward online credit card use. In this context, we aim toanswer the following research questions: (1) How are consumers’negative attitudes changed by different levels of message proxim-ity? (2) How is the effect of message proximity on credit card usageattitudes moderated by the message source? (3) Is the answer to re-search question 1 the same for positive-attitude consumers?

The next section introduces the conceptual and theoretical foun-dation of the study, followed by a discussion of the hypotheses. Theexperimental procedure and methods of data analysis are describedin Sections 3 and 4. Finally, in Section 5 we discuss the findings andtheir implications, as well as the limitations of the study.

2. Conceptual framework

2.1. Social judgment theory

Social judgment theory is a theory of persuasion (Sherif andHovland 1980). It defines attitude change as ‘‘the stands the indi-vidual upholds and cherishes about objects, issues, persons,groups, or institutions’’ (Sherif et al. 1965). The essence of the the-ory is that the degree of attitude change caused by a persuasivemessage depends on how the recipient evaluates that message.According to the theory, attitude change is a two-step process.First, individuals who hold a position on an issue indicate this po-sition by marking an anchor point on an attitude continuum rang-ing from acceptance to no commitment to rejection. Where themark is placed dictates the relative likelihood of assimilation andcontrast. One of these two effects occurs when a discrepant view-point is expressed in the message. The closer the message is to theanchor point, the more likely it is to be assimilated; the more dis-tant it is, the more likely it is to be contrasted. Thus, the secondstep involves the message recipient mentally measuring the dis-tance between the anchor point and where the message falls onthe continuum. If the message is dissuasive and the recipient’s an-chor point (ex ante attitude) represents rejection of the message,assimilation occurs. If the message is persuasive but the recipient’sanchor point is in the rejection region, contrast occurs. Thus, if thedistance between the two points is too little or too much, attitudechange is unlikely. On the other hand, attitude change is likely ifthe recipient’s anchor point is placed in the no-commitment regionand the message is in either the acceptance or rejection region, or ifthe message falls in the no-commitment region and the recipient’sex-ante attitude is in the region of acceptance or rejection.

Social judgment theory has been widely used in domains suchas interpersonal communication (Sherif and Hovland 1980), orga-nization policy announcements (Wood and Bandura 1989),government publicity (Krosnick 1990), and health education(Heald 1991). In interpersonal communication, for example, Berndt(1976) applied social judgment theory to investigate interpersonalconflicts caused by cognitive dissonance. He concluded that it ispossible for two persons with divergent thoughts to reach agree-ment if the researcher applies social judgment theory to evaluatethe discrepancy between their positions. Accordingly, the socialjudgment theory is applicable to our study because marketersoften advocate their own trust-assuring messages, whereas con-sumers with a negative attitude toward credit cards usually findsuch messages suspect. Social judgment theory helps us under-stand how consumers change their positions after being con-fronted with a message arguing for a different position.

Although social judgment theory is useful for understandingattitude change, its focus is exclusively on evaluating the distancebetween the individual’s anchor point and the location of the mes-sage; it is not on the message’s content per se (Siero and Doosje1993). In addition, according to Chaiken and Maheswaran (1994),

recipients’ decisions are susceptible to messages endorsed by acredible source. However, social judgment theory provides noway to empirically test the effect of source credibility (Heesackeret al. 1983). Our study was intended to fill this gap by examiningthe effects of messages varying in source as well as content.

2.2. Cialdini’s principles of persuasion

Social judgment theory allows us to understand the process ofpersuasion, but if we want to completely understand how persua-sion works we need to identify the factors that affect it. Thesefactors and their role in the persuasion process can be graspedby applying Cialdini’s (1993) Principles of Persuasion. Cialdiniproposed seven principles that persuaders can use to improvethe persuasiveness of their messages: reciprocity, commitment,liking, social proof, authority, contrast, and scarcity.

Reciprocity refers to the feeling of indebtedness that comes froma persuader’s generosity, manifested through vehicles such as gifts,concessions, and favors. Commitment represents the fact that oncea persuader and a recipient are committed to each other, they bothhave a strong desire to maintain that commitment. Liking meansthat recipients’ decisions or behavior are more easily influencedby persuaders they are favorably disposed to than those they areunfavorably disposed to. Social proof, or social conformity, meansthat recipients usually must draw on the experiences of others tomake difficult decisions. Although these four principles have beenshown to be valid, our research did not address the effects of mu-tual benefits (reciprocity), commitments by both parties (commit-ment), feelings toward specific persuaders (liking), or socialrecognition (social proof) based on trustmarks. Instead, we wereinterested in the source and proximity of the message, which areaddressed by the other three of Cialdini’s principles.

The principle of authority is the extent to which a persuader hasa certain level of prestige, reputation, or expertise. Recipients gen-erally believe that accepting information from a persuader withrecognized authority decreases the probability of making a wrongdecision. The authority principle can be represented by trustmarksthat reflect the prestige, reputation, or expertise of the persuader.Contrast refers to information that can be used to distinguish thepros and cons of whatever is being compared. For example, if peo-ple first lift a heavy stone and then a light stone, they will feel thesecond stone to be lighter than they would had they not lifted theheavy stone first. Thus, applying the contrast principle in trust-marks can inform consumers that the website has better securitythan competing sites. Please note that contrast in Cialdini’s princi-ples is different from contrast in social judgment theory. ForCialdini, scarcity refers to objects of persuasion that are bothvaluable and rare. It has been shown that objects and opportunitiesbecome more valuable as they become less available (Mazis 1975).The scarcity principle explains why trust-promoting protectionmechanisms are safe: they are seldom descrambled by hackers.

2.3. Framework

In Fig. 1, the black line in bold refers to the attitude regions onthe continuum, including rejection, no commitment, and accep-tance. The issue at hand is what consumers with a negative atti-tude toward online credit cards do when confronted with a trust-promoting message. Initially, they anchor their ex ante attitude inthe rejection region. Second, when they receive the trust-promot-ing message, they evaluate the distance between the anchor pointand the region where the message falls. According to the socialjudgment theory, if a consumer is being persuaded to accept amessage landing inside the region of rejection (messages with highproximity: authority only), the fact that it is so close to the anchor

Fig. 1. Conceptual framework for consumers with ex ante negative attitudes.

W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345 337

point results in the ex post attitude remaining the same as the exante attitude; this is assimilation.

The other effect is contrast. If consumers are persuaded toaccept a message that is far away from their anchor point(messages with low proximity: authority-plus-contrast), they areless likely to change their attitude, according to social judgmenttheory. A moderate distance between the anchor point and themessage point (messages with moderate proximity: authority-plus-contrast-plus-scarcity) leads more easily to attitude changethan do assimilation and contrast.

We maintain that consumers with negative attitudes are notlikely to change these attitudes when high proximity or low prox-imity messages are received, because high proximity messagesprovide too little information and low proximity messages makepromises with no additional explanation to back up the theory.In principle, moderate proximity messages should be the most per-suasive, because they include additional explanation to buttressthe promise. This leads consumers to place their mark in the regionof no commitment, not too close or too far from their anchor posi-tion. In this case, the likelihood of attitude change is maximized.

2.4. Research model and hypotheses

Based on our integration of Cialdini’s principles with socialjudgment theory, we used three types of persuasive messages,operationally defined as the psychological distance between themessage content and the consumer’s ex ante attitude. (Batra andStayman 1990, Kim and Benbasat 2009, Sachse and Gierl 2009):

� Authority-only (high proximity): the trustmarks were limited toauthority elements and the credibility associated with them,without explanation (Maison et al. 2004).� Authority-plus-contrast (low proximity): in addition to providing

trustmarks, the website conveyed a message to informed con-sumers that it has better security than other websites theymight visit.� Authority-plus-contrast-plus-scarcity (moderate proximity): in

addition to authority and contrast, the website included a sec-ond trust-promoting message that told users that the websitewas so safe that it would take a trillion years to decode thecredit card information.

This is based on Toulmin’s (2003) argument that the structureof a message has three sequential or additive components: claim,data, and backing. The claim is the assertion being put forwardfor acceptance. The data comprise the information on which theclaim is based. The backing is the evidence that supposedly justifies

acceptance of the data. In our experiment, the claim said that thetransaction is secured 100% by the trustmark. The supporting dataconsists of the ‘‘fact’’ that ‘‘our longer key is better (i.e., 128 bit vs.256 bit).’’ The backing provided the evidence for the validity of thedata: ‘‘It takes an unreasonable amount of time to unscramble ourkey.’’ This claim-data-backing scenario corresponds to Cialdin’sauthority-plus-contrast-plus-scarcity principle. The sequentialnature of this principle also explains why we did not simplypropose authority-plus-scarcity.

We expected that the moderately proximate message would bethe most persuasive for those who did not use a credit card online.The first two hypotheses are:

Hypothesis 1 (The Authority-Plus-Contrast-Plus-Scarcity Hypothe-sis). On a website, trust-promoting messages that are character-ized by authority, contrast, and scarcity create more positiveattitudes toward credit cards than trust-promoting messagescharacterized by authority and contrast alone.

Hypothesis 2 (The Authority-Plus-Contrast Hypothesis). On a web-site, trust-promoting messages that are characterized by authorityand contrast create more positive attitudes toward credit cardsthan trust-promoting messages that are characterized by authorityalone.

Not all trustmark issuers are equally credible to consumers,even though they handle information security in the same way.Source credibility is defined as the extent to which an informationsource is perceived to be believable, competent, and trustworthyby information receivers (Hovland and Weiss 1951). Odom et al.(2002) also demonstrated that the process by which trustmarksinfluence consumers’ online purchase behavior involved recogni-tion of and familiarity with a particular trustmark. Because issuinga trustmark is costly, it is necessary to identify which trustmark ismost easily recognized by consumers, and then use it instead ofmultiple trustmarks.

Following this prescription, we identified two message re-sources: international and domestic, and tested their moderating ef-fects on message proximity. A trustmark issued by an internationalissuer is the one frequently featured on the shopping websites, whilea trustmark issued by a domestic issuer is relatively new and cannoteasily capitalize consumers’ impressions. We propose:

Hypothesis 3 (The Domestic vs. International Trustmarks Hypothe-sis). Because the effect of message proximity on attitudes towardcredit cards is moderated by trustmarks issued by different thirdparties, international trustmarks are more persuasiveness thandomestic trustmarks.

Fig. 2. Research model.

338 W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345

We test two things. The first is how message proximity affectsconsumers’ attitudes toward using credit cards. The second is themoderating effect of the message source. The resulting researchmodel is shown in Fig. 2.

3. Research procedure

3.1. Participant selection

Since our goal is to find the best way to change consumers’ neg-ative attitudes toward online credit card use, it is necessary toscreen eligible participants with both negative and positive creditcard attitudes and then make a comparison between them. Weadministered a preliminary questionnaire that was being pro-moted on three websites popular in Taiwan: PTT, Facebook, andPlurk. The questionnaire consisted of the following items: Haveyou had any online shopping experience before? Do you have acredit card issued by a bank? Have you ever used a credit card toshop online, and, if not, what is the most frequent way you checkout online? What kind of merchandise have you purchased online,and what reasons are there that you do not use a credit card whenyou shop online? What products are you most likely to buy onlineif the concerns you listed above are eliminated?

Respondents who answered Questions 1 and 2 with ‘‘Yes,’’Question 3 with ‘‘No,’’ and Question 5 with ‘‘did not use creditcards online because of security concerns’’ were classified as hav-ing a negative attitude toward using a credit card online. Respon-dents who answered Questions 1 through 3 with ‘‘Yes’’ wereclassified as having a positive attitude toward credit cards. Respon-dents who answered Questions 1 or 2 with ‘‘No’’ were disqualifiedbecause they either did not have credit cards or they did not shoponline. In either case, they belonged to a population that is notwithin our research domain. Questions 5 and 6 were only scoredif Question 3 was answered with ‘‘No.’’

Because domain knowledge about the trustmark might influ-ence acceptance of the persuasion, we controlled for such knowl-edge by asking participants to rate the statement ‘‘The SSLsecurity protection is provided by a shopping website rather thana third party’’, using a five-point Likert scale from 1 to 5, where 1means ‘‘strongly agree’’ and 5 means ‘‘strongly disagree.’’ Wewanted to select participants whose domain knowledge was belowthe median on this scale. This filtering procedure, which is similarto the one used by Wang and Benbasat (2009), mitigated thepotential confounding effect of domain knowledge. Of the 689respondents to the original survey, 69 stated that they shopped on-line but did not have credit cards, and 149 scored above the med-

ian on trustmark knowledge. These 218 respondents (32%) wereexcluded from the study. Of the remaining 471, 177 (26%) statedthat they owned a credit card but did not use it for onlinepayments due to security concerns. Finally, after removing 20respondents who did not want to participate further, we were leftwith a sample of 120 eligible participants with negative attitudestoward credit cards. Of the 689 original respondents, 294 statedthat they used credit cards online (42%). However, 33 of thesedid not want to participate further, leaving a sample of 261 partic-ipants with positive attitudes toward credit cards to serve as thecontrol group.

3.2. Experimental design

To test our hypotheses, we combined a survey with an experi-ment. The latter incorporated a 3 � 2 factorial design with threemessage proximity levels and two message source levels. We chosethe TWCA (domestic in Taiwan) and VeriSign (international) logosfor our study. TWCA is relatively new and less well known thanVeriSign. Next, we assigned experiments to participants, and com-pared the pre-message and post-message results.

Participants were divided into six groups, each of which re-ceived a different treatment. (See Table 1.) An email was sent toeach of the participants to explain what they needed to do. First,they were requested to complete the pre-message attitude ques-tionnaire, which they were told would be used to check their atti-tude change after they were given the trustmark messages. Second,they needed to visit the virtual shopping website illustrated inFig. 3. The website contained information about the products,including pictures, prices, and brief explanations of the products’functions.

Cameras were chosen as the target product based on two crite-ria. The first was the answer to Question 6 in the preliminary ques-tionnaire. More than half of the respondents (57%) stated that theywould feel safer buying electronic products if online shopping wasprotected. The most popular product choice was a camera. Second,we wanted a high-price product, because we expected that themore participants had to pay for the product, the more susceptiblethey would be to the trust-promoting message.

Third, participants needed to consummate the purchase with acredit card. Their decisions were to be based on the informationthey received at check-out, which varied by treatment in the3 � 2 design. (See Table 1 again.) For example, the check-out pagefor Group 1 showed trust-promoting messages that contained onlyauthority information along with the domestic trustmark. (SeeFig. 4.)

Table 1Experimental group and trust-promoting messages.

Message proximity Message source

Domestic trustmark International trustmark

High proximity(authority-only)

Group 1 Group 2

Low proximity(authority-plus-contrast)

In contrast to other shopping websites that use only 128 or 256 bitsof SSL security encryption, our website uses a world-high 1024 bitSSL to assure the safety of your credit card information. We arededicated to providing our customers with a safe and reliable onlinetransaction environment

In contrast to other shopping websites that provide only 128 or 256bits of SSL security encryption, our website uses a world-high 1024bit SSL to assure the safety of your credit card information. We arededicated to providing our customers with a safe and reliable onlinetransaction environment

Group 3 Group 4

Moderate proximity(authority-plus-contrast-plus-scarcity) In contrast to other shopping websites that use only 128 or 256 bits

of SSL security encryption, our website uses a world-high 1024 bitSSL to assure the safety of your credit card information. We arededicated to providing our customers with a safe and reliable onlinetransaction environment

In contrast to other shopping websites that use only 128 or 256 bitsof SSL security encryption, our website uses a world-high 1024 bitSSL to assure the safety of your credit card information. We arededicated to providing our customers with a safe and reliable onlinetransaction environment

SSL encryption ensures that information is scrambled in transit sothat only the authorized recipient can decode it. The number of bitstells you the size of the key. At current computing speeds, a hackerwith the motivation to attack would require a trillion years to breakinto a session with 1024 bit encryption. In other words, it isimpossible for our data to be stolen in a reasonable time. So you canfeel free to leave your credit card information on our website

SSL encryption ensures that information is scrambled in transit sothat only the authorized recipient can decode it. The number of bitstells you the size of the key. At current computing speeds, a hackerwith the motivation to attack would require a trillion years to breakinto a session with 1024 bit encryption. In other words, it isimpossible for our data to be stolen in a reasonable time. So you canfeel free to leave your credit card information on our website

Group 5 Group 6

Fig. 3. The main page of experimental shopping website.

W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345 339

To ensure that the participants paid attention to the message,they were requested to complete a short questionnaire that askedthem to choose the correct message (which they had just seen) in

each of the six treatment conditions. They could not go to the nextpage until they did so. Finally, the participants were presentedwith the post-message attitude and demographic questionnaires.

Fig. 4. Check-out page on the experimental shopping website.

Table 2Summary of measurement items.

Construct Measure Loading AVE

Pre-message attitude (PRA)composite reliability = 0.94PRA1 The main reason that I do not use a credit card to shop online is transaction security. 0.85 0.80PRA2 The transaction security mechanisms provided by the website are not strong enough. 0.91PRA3 To me, using a credit card to shop online is dangerous. 0.92PRA4 For my safety, I do not prefer credit cards as my online payment method. 0.91

Post-message attitude (POA)composite reliability = 0.93POA 1 The transaction security guaranteed by the website increases my confidence in using a credit card online. 0.88 0.79POA 2 The transaction security provided by the website eases my worries about using a credit card online. 0.88POA 3 The data encryption provided by the website effectively reduces my concern about using a credit card online. 0.91POA 4 After reading the trust messages provided by the website, I would like to use a credit card to shop online. 0.88

340 W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345

All participants received a $10 reward for their participation. Tofurther motivate them to concentrate on the message, they weretold before the experiment that one of them would be selectedby lottery to win an Android Pad (e-reading device) following ful-fillment of the experimental requirements.

3.3. Attitude measurement

As shown in Table 2, both pre-message and post-message atti-tudes were measured using four 5-point Likert scale items adaptedwith minor modifications from Joo et al. (2003). For pre-messageattitudes, 1 meant ‘‘strongly agree’’ and 5 meant ‘‘strongly dis-agree.’’ For post-message attitudes, 1 meant ‘‘strongly disagree’’and 5 meant ‘‘strongly agree.’’ Prior to the formal analysis, preli-minary data examination was conducted. We screened for missingdata, outliers, construct reliability, and construct validity. No miss-ing data or outliers were found. Reliability was evaluated usingcomposite values. Hair et al. (2006) recommend an acceptance le-vel of 0.7 for composite reliability. As summarized in Table 2, the

composite reliabilities of our two attitude constructs exceeded0.93, this meeting the criterion.

Fornell and Larcker (1981) suggested two criteria that should bemet to establish convergent validity. First, all the factor loadingsshould not only be significant but also exceed 0.5. Second, the aver-age variance extracted (AVE) for each construct should exceed themeasurement error variance for that construct (AVE should begreater than 0.5). All the items listed in Table 2 exhibit loadingsgreater than 0.85 within their respective constructs and all AVEsare larger than the error variance. Thus, both criteria for conver-gent validity were met. Discriminant validity is the extent to whicha construct and its indicator variables differ from another constructand its indicator variables (Bagozzi and Phillips 1991). We exam-ined it using a criterion suggested by Fornell and Larcker (1981):the square root of the AVEs should be greater than the correlationbetween the two constructs. Table 3 shows that the correlation be-tween the pair of constructs was less than the corresponding AVEs(diagonal values). All the constructs met the requirement, provid-ing evidence of discriminant validity.

Table 3Discriminant validity analysis.

Construct AVE PRA POA

Pre-message attitude (PRA) 0.80 0.89Post-message attitude (POA) 0.79 �0.57 0.89

Diagonal elements (in bold) are the square roots of the average variance extracted(AVE). Off-diagonal elements are the correlations among constructs.

Table 4Message difference check.

Message proximity (ANOVA test) Message source (t-test)

High Low Moderate Domestic InternationalMean = 1.78 Mean = 2.50 Mean = 4.18 Mean = 2.03 Mean = 3.62

F(2,117) = 151.68,p < 0.01

t = 13.95,df = 118,p < 0.01

Table 5Demographics.

Measure Items Frequency Percentage

Gender Male 62 51.7Female 58 48.3

Age Under 20 3 2.520–25 40 33.326–30 51 42.5Over 30 26 21.7

Occupation Officeworker

65 54.2

Student 30 25.0Manuallaborer

14 11.7

Other 11 9.1

Online shopping experience(current year)

1 time 10 8.3

2 times 47 39.23 times 33 27.54 times 18 15.0>4 times 12 10.0

Table 6Two-way ANOVA on attitudes in all conditions.

Source SS (Type III) Df MS F

Intercept 1164.076 1 1164.076 3525.569*

Message proximity 36.570 2 18.285 55.378*

Message source 15.230 1 15.230 46.125*

Proximity � source 0.547 2 0.273 0.828Error 44.841 114 0.393Corrected total 88.145 119

* p < 0.01.

W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345 341

4. Analysis and results

4.1. Message manipulation check

To check that message proximity and message source wereviewed differently by the participants, we asked them the follow-ing two questions after they answered the post-message attitudequestionnaire: To assess message proximity, we asked ‘‘How muchtrust-promoting information was available on the check-out page’’(Parra and Ruiz 2009); to assess source familiarity, we asked themto rate the statement ‘‘I am familiar with the trustmark providedby the website’’ (Komiak and Benbasat 2006). Both questions wereanswered on a five-point Likert scale from 1 to 5, where 1 means‘‘strongly disagree (very little)’’ and 5 means ‘‘strongly agree (verymuch).’’ Results showed that message proximity was successfullydifferentiated. (See Table 4.) Participants indicated that more infor-mation was available on the website in the moderate proximitycondition than in the low and high proximity conditions. Partici-pants also indicated that the message sources were significantlydifferent in terms of perceived familiarity.

In addition, because we used a self-report survey to evaluatepre-message and post-message attitudes, it was necessary to checkfor the potential impact of common method variance, which weaccomplished through Harman’s single-factor test (Podsakoffet al. 2003). If substantial common method variance is present,either a single factor will emerge from an exploratory factor anal-ysis or one general factor will account for the majority of thecovariance among the constructs. The unrotated factor analysisrevealed the presence of eight attitude items with eigenvaluesgreater than 1.0, but no general factor. The eight items accountfor 77% of the total variance. The first factor explains only 40% ofthe variance, indicating no apparent common method variance.

4.2. Demographics

Table 5 summarizes the participant characteristics. Of the 120participants with negative attitudes toward credit cards, 51.7%were male and 54.2% were office workers. Most (75.8%) were inthe 20 to 30 year age group, consistent with Taiwan’s legal require-ment that credit card holders be at least 20 years old. Nonetheless,we found three credit card holders who were underage. Becausethey may have been using a card associated with the primary card,we did not remove them from the analyses. In regard to onlineshopping experiences, 39.2% of the participants stated that theyhad made at least two purchases on the Internet, whereas 27.5%cited three purchases.

4.3. Tests of hypotheses

To test the hypotheses, we performed a two-way ANOVA. (SeeTable 6.) The results revealed that both message proximity(F = 55.37, p < 0.01) and message source (F = 46.13, p < 0.01) signif-icantly affected post-message attitudes, but the interaction be-tween message proximity and message source was notsignificant (F = 0.828). For this reason, separate one-way ANOVAswere conducted for the two treatments (Hair et al. 2006).

Table 7 shows that post-message attitudes differed significantlyacross the message proximity groups (F = 40.05, p < 0.01). Multiplecomparisons performed with the Tukey HSD test revealed that theauthority-plus-contrast-plus-scarcity condition was associatedwith significantly more positive attitudes than the other two condi-tions. (See Table 8.) Thus, the Authority-Plus-Contrast-Plus-ScarcityHypothesis (H1) was supported. However, no significant differencewas found between the authority- only and authority-plus-contrastconditions; thus, the Authority-Plus-Contrast Hypothesis (H2) was

rejected. After the criterion p-value was made more stringent (froma = 0.05 to a = 0.01), Bonferroni tests produced results similar to thepreviously reported Tukey results, decreasing the chance of makinga Type I error to very acceptable levels.

Fig. 5 gives the analysis of the change from pre-message andpost-message attitudes for participants with negative pre-messageattitudes toward credit cards. In all six conditions, these attitudesbecame significantly more positive after participants read themessage. The change was greatest in the authority-plus-contrast-plus-scarcity condition. However, there was no significant differ-ence between the domestic and international trustmarks(F = 0.828); thus, the Domestic vs. International TrustmarksHypothesis (H3) was rejected. (See Table 6 again.)

Table 7One-way ANOVA on attitudes in relation to message proximity.

Source SS (Type III) Df MS F Homogeneity of variance

Intercept 1164.076 1 1164.076 2549.682* F = 2.837Message proximity 36.570 2 18.285 40.050* df1 = 2Error 53.417 117 0.457 df2 = 117Corrected total 89.987 119 p = 0.063.

* p < 0.01.

Table 8Multiple comparisons (H1, H2).

Turkey HSD condition (I) Comparison condition (J) Mean difference (I � J) 95% confidence interval

Lower bound Upper bound

1. Authority-only (Mean: 2.58) 2 �0.30 �0.66 0.523 �1.29* �1.65 �0.94

2. Authority- Plus-Contrast (Mean: 2.89) 1 0.30 �0.52 0.663 �0.99* �1.34 �0.63

3. Authority- Plus-Contrast- Plus-Scarcity (Mean: 3.88) 1 1.29* 0.94 1.652 0.98* 0.63 1.35

Bonferroni-adjusted condition (I) Comparison condition (J) Mean difference (I � J) 99% confidence interval

Lower bound Upper bound

1. Authority-only (Mean: 2.58) 2 �0.30 �0.69 0.083 �1.29* �1.68 �0.91

2. Authority- Plus-Contrast (Mean: 2.89) 1 0.30 �0.08 0.693 �0.99* �1.37 �0.60

3. Authority-Plus- Contrast-Plus-Scarcity (Mean: 3.88) 1 1.29* 0.91 1.682 0.98* 0.60 1.37

Note: The standard error for all analyses is 0.15.* p < 0.01.

Fig. 5. Mean attitude change (pre-message to post-message) in the negative group. Fig. 6. Mean attitude change (premessage to postmessage) in the positive group.

342 W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345

We repeated the attitude change analysis on data from the 261participants who entered the study with a positive attitude aboutcredit cards. Fig. 6 shows that in the authority-only condition,these participants’ attitudes became significantly more negative.

This finding is similar to that reported by Karahanna et al.(1999), who found that when users gained experience with thesystem, usefulness became more important to them than ease ofuse. The fact that there was no significant attitude change in theother two conditions (authority-plus-contrast and authority-plus-

W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345 343

contrast-plus-scarcity) for either trustmark suggests that consum-ers with positive credit card attitudes expect to see more than justan authority message on the websites they visit. This resultmatched the results we obtained for the participants with negativeattitudes. For the positive-attitude participants, authority-plus-contrast and authority-plus-contrast-plus-scarcity had highproximity, because these messages confirmed that online creditcard transactions had good security features and thus their deci-sion to use the card online was correct. The authority-only mes-sages could have had moderate proximity because they mighthave conveyed the new implication that the websites did not havethe security features that the participants expected.

5. Discussion

The invention of credit cards and online transactions are twomilestones in financial services history. They have affected eco-nomic activity, and their promotion has helped to make economiesmore prosperous. As many people still do not trust online creditcards though, changing their attitudes is essential. This study isone of the first to investigate the effects of the source and proxim-ity of messages on consumers’ negative attitudes toward using acredit card online. The three levels of message proximity used inthe experiment were intended to assess their different effects onpositive change in consumers’ credit card attitudes. The author-ity-plus-contrast-plus scarcity presentation was the most effectivein creating more positive attitudes toward credit card use, followedby the authority-plus-contrast and authority-only presentations.

However, it is worthwhile to note that the Authority and Con-trast Hypothesis (H2) was rejected: although the authority-plus-contrast condition produced more attitude improvement thanthe authority-only condition, the difference was not significant. Aplausible reason for this outcome may be that the participants witha negative attitude toward credit card use did not perceive the dif-ferences between the two contrast messages (i.e., 128 bit vs. 256bit SSL). Thus, they could not understand how difficult it wouldbe for hackers to deactivate the SSL encryption. For these partici-pants, 256 bits is just twice 128 bits, not an exponential increment.That is why the authority-plus-contrast condition created no moreattitude improvement than the authority-only condition. In theauthority-plus-contrast-plus-scarcity condition, the scarcity mes-sage further explained how difficult it would be for hackers todeactivate the SSL encryption (it would require a trillion yearsfor them to break into data with 256 bit encryption). This resultsuggests that consumers with negative credit card attitudes arepersuaded only if they are provided with a moderately proximatemessage (authority-plus-contrast-plus-scarcity).

The most puzzling result in this study was the lack of a signifi-cant interaction between message proximity and message source(the Domestic vs. International Trustmarks Hypothesis, H3). Thisfinding suggests that message proximity plays a major role inimproving credit card attitudes, regardless of whether the con-sumer recognizes the trustmark. As a result, consumers with a neg-ative credit card attitude are unlikely to be affected by the sourceof the trustmark, because they have no experience in using creditcards online. Our finding in this regard is similar to that of Kovaret al. (2000), who found that trustmarks had a positive influenceon consumers’ attitudes only when their attention was drawn tothem. Trustmark understandability may also explain the lack of asignificant interaction between message proximity and messagesource. Aiken and Boush (2006) argued that the presence of a trust-mark has a greater effect on inexperienced than experienced con-sumers, but the magnitude of the effect depends on how wellconsumers with a negative credit card attitude understand thetrustmark. For these authors, the source effect of the trustmark is

at best only a graphical difference, with no implications for thedifferent effects of different trustmarks. In this case, consumers’attitudes toward using a credit card online may not be affectedby how they perceive trustmarks. Thus, an additional trust-assuring message aimed at increasing the understanding oftrustmarks is needed. Nöteberg et al. (2003) suggested in thisconnection that consumers with adequate knowledge about atrustmark are more likely to notice it on a shopping website thanthose without such knowledge.

5.1. Theoretical implications

First, unlike prior research that focused mainly on whether ornot to have a trustmark on one’s website (Head and Hassanein2002, Kimery and McCord 2002, Lala et al. 2002, Miyazaki andKrishnamurthy 2002, Cook and Luo 2003, Zhang 2004), this studyexplored the extent to which trust-assuring messages can be usedto improve consumers’ attitudes toward using a credit card online.Nowadays, most shopping websites use trustmarks to persuadeconsumers that the website they visit is safe. However, as evi-denced by the survey mentioned in the introduction section (MIC2010), merely having a trustmark on one’s website does not neces-sarily guarantee consumers’ patronage, especially if they have anegative attitude toward credit cards. Therefore, our research find-ings are relevant to businesses that want to leverage consumers’favorable perceptions of their website’s credit card transactionsso as to attract more patronage.

Second, the present study provides empirical evidence that so-cial judgment theory can be fortified by combining it with Cialdi-ni’s model to further our understanding of online shopping.Social judgment theory emphasizes the discrepancy betweenpre-existing attitudes and the position advocated in the message,while paying no attention to the content of the message. Specifi-cally, we applied the three principles in Cialdini’s model (authority,contrast, and scarcity) to increase the predictive power of socialjudgment theory. The findings suggest that consumers’ negativecredit card attitudes can be improved if the consumers are pro-vided with moderately proximate trust-assuring messages. Howfavorable a consumer’s attitude is toward a website generally de-pends on the quality of the website’s structure and informationcontent (DeLone and McLean 2003). Cialdini’s model can help withthe design of high quality and persuasive trust-assuring messages(information), whereas the social justice theory addresses the atti-tude change created by these messages.

Third, our study has established that the source effect of a trust-mark is not a moderator of the relationship between message prox-imity and post-message attitude. The domestic trustmark hadabout the same effect as the international trustmark on attitudeimprovement. Unlike prior studies advocating the impact of mes-sage source on attitude change, ours suggests a new perspective,namely, that the source of the trustmark used in a persuasive mes-sage is irrelevant to consumers with a negative attitude towardcredit card use.

In sum, our study provides evidence that different trust-assur-ing messages can have dramatically different effects on attitudeimprovement, depending on their consistency with consumers’pre-existing attitudes toward using a credit card online. We con-clude that when trust-assuring messages are moderately proxi-mate (authority-plus-contrast-plus-scarcity), consumers are verylikely to change their credit card attitudes from predominatelynegative to predominately positive.

5.2. Practical implications

Transaction security has been found to influence consumers’attitudes toward shopping online (Liao and Cheung 2001). This

344 W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345

finding is consistent with the proposition of Chase and Dasu (2001)that service encounters should end satisfactorily and that negativeexperiences should be overcome as soon as possible. This is be-cause events (e.g., expressions of security concerns) that occur nearthe end of an encounter (e.g., checkout stage) have a moreprofound impact than earlier events on the overall shopping expe-rience. Therefore, proprietors must pay attention to their trust-assuring message as soon as consumers give them their credit cardnumbers, particularly if those consumers have a negative attitudetoward credit cards.

Drawing on our findings, we provide several specific recom-mendations for practitioners. Our data reveal that consumers’ atti-tudes toward using a credit card online can be affected by thepersuasive messages they receive. Providing consumers havingnegative credit card attitudes with additional messages that meettheir security concerns can facilitate attitude improvement. A via-ble way to achieve this goal is to provide security-strengtheningmessages on the website’s checkout page, including a trustmarkthat consumers can understand. This approach gives consumersthe opportunity to reassess their pre-existing negative attitudestoward incoming persuasive messages, creating the potential forattitude improvement.

Our results suggest that the effect of message proximity onpost-message attitudes is not affected by the source of the trust-mark, indicating that the message source is probably a necessarybut insufficient condition for attitude improvement. Thus, we rec-ommend that proprietors provide additional messages to expandconsumers’ understanding of why a trustmark is important andhow it operates. The findings also suggest that trust-assuring mes-sages lead to the greatest attitude improvement if they containauthority-plus-contrast-plus-scarcity information. Regardless ofmessage source, authority-only information gives consumers noreason to accept the persuasive message, and thus it produces verylittle attitude improvement. This suggestion seems straightfor-ward, but the trustmarks used by most websites contain authorityinformation only. Thus, we recommend that proprietors give theirconsumers moderately proximate messages, because consumersgenerally feel that a trustmark is just a symbol, and they lackany specific understanding of its meaning (Williams and Grimes2010). Logically, the more deficient the explanatory capacity of amessage, the less persuasive it is (Smith et al. 1991).

The insignificant interaction between message proximity andmessage source suggests that when consumers intend to purchasea high-price product (as was the case in our experiment), particu-larly if they have a negative attitude toward credit card usage, per-suasion by an independent, third-party source does not necessarilyimprove their attitudes; rather, consumers form their attitudes byexamining the persuasive messages provided on websites. Thissuggests that both domestic and international trustmarks areeffective if they are explained by additional messages. Someshopping websites license two or more trustmarks with the samepurpose, because they believe that the more cues on the website,the more persuasive the message is to consumers. However, thisapproach is not cost-effective. We therefore suggest that propri-etors choose either a domestic or an international trustmark,whichever is cheaper, and feature it on the website along with amoderately proximate message.

5.3. Limitations

The present study has certain limitations. First, there wereproblems with our sample of participants. We were unable to ob-tain as many participants with a negative attitude toward using acredit card online as we would have liked. Of the 689 respondents,only 120 met this criterion, and they were randomly and equally

assigned to the six experimental treatments (n = 20). Although thisgroup size meets the minimum requirement for multivariate anal-ysis (Hair et al. 2006), larger samples would have increased statis-tical power. To control for this sampling bias, we added theanalysis of data from participants with positive attitudes towardcredit card use. The results led us to conclude that the messagemanipulation had a positive effect on both the positive-attitudeand negative-attitude groups.

A second sampling issue impacts the generalizability of ourfindings. When we conducted our message source comparison,we found that VeriSign was more persuasive than Taiwan’s TWCAbut the result is not statistically significant. If we want to concludethat international trustmarks are more persuasive than domesticones, we may need to sample other countries in addition toTaiwan. A third potential sampling bias is related to the fact thatwe selected only people who own credit cards and shop online.However, the fact is that there are people who own credit cardsbut do not shop online at all because they distrust online shopping.Such persons, who would fall in the negative attitude category,were not represented our study. We decided against sampling thissubpopulation because to do so would have further biased thesample. We were afraid that some individuals of this type wouldrefuse to participate because we were asking them to change froman online non-shopper to an online shopper. The problem is thatthose willing to participate may have had less security concernsthan those who refused to participate, thereby not properly repre-senting the attitudes of this subpopulation. Thus, to fully under-stand how message manipulation affects people manifestingdifferent kinds of credit card or online shopping behavior, wemay need a more comprehensive sampling approach that stratifiesdifferent population subgroups based on multiple criteria such asshopping online, credit card use, or security concerns.

Second, because we used only higher-priced products (digitalcameras), our findings may not generalize to lower-priced prod-ucts. Although consumers have been shown to be less sensitiveto lower-priced products (Ba and Pavlou 2002), it does not followthat trust-promoting mechanisms are unimportant for such prod-ucts. In fact, consumers who do not take advantage of the availablesecurity measures when using credit cards online may find thattheir financial losses are greater if they conduct many small trans-actions than if they conduct a few large transactions. Thus, we planto conduct a similar study using lower-price products.

Third, following the academic paradigm, we based our hypoth-eses on existing theory: Toulmin’s additive message structure,social justice theory and Cialdini’s principles. However, there isno theory stating that other combinations of authority, contrast,and scarcity are not valid. Two research questions are particularlyimportant. What would be the effect of sequencing the message asauthority ? scarcity ? contrast instead of authority ? contrast ?scarcity? How would combining authority with one of the othertwo types affect attitude change? Such a study might help us com-pare the different effects of contrast and scarcity. However, suchresults could still be hard to interpret.

Psychologically speaking, authority-plus-scarcity may be per-suasive enough without contrast; whereas a scarcity statementthat states, for example, that trillions of years are required to de-code an encryption is clear and strong, a statement that says thenumber of bytes needed for encryption has been doubled from128 to 256 is not likely to impress people who do not know muchabout information technology. On the other hand, social justicetheory predicts just the opposite: if the distance between the mes-sage recipient’s anchor point and the message proximity point istoo large or too small, attitude change is unlikely. There are manyother possible combinations. For either of the above two issues, weneed a between-subjects design, because the results would be

W. Shu, C.Y. Cheng / Electronic Commerce Research and Applications 11 (2012) 335–345 345

contaminated if the same participants were exposed to differentmessage structures. Thus, a full study of all possible combinationsis likely to require many more participants.

To conclude, security concerns may not be the only reason con-sumers refrain from using credit cards online. For example, theymay be concerned about the confidentiality of their identity infor-mation. Thus, our final recommendation is that future studies ex-plore these other reasons, using redesigned trust-promotingmessages.

References

Aiken, K. D., and Boush, D. M. Trustmarks, objective-source ratings, and impliedinvestments in advertising: investigating online trust and the context-specificnature of Internet signals. Journal of the Academy of Marketing Science, 34, 3,2006, 308–323.

Aiken, K. D., Osland, G., Liu, B., and Mackoy, R. Developing Internet consumer trust:exploring trustmarks as third-party signals. In R. Geraldine (ed.), MarketingTheory and Applications, American Marketing Association, Henderson andMarian Chapman Moore, Chicago, IL, 2003, 145–146.

Ba, S., and Pavlou, P. A. Evidence of the effect of trust building technology inelectronic markets: price premiums and buyer behavior. MIS Quarterly, 26, 3,2002, 243–268.

Bagozzi, R., and Phillips, L. Assessing construct validity in organizational research.Administrative Science Quarterly, 36, 3, 1991, 421–458.

Batra, R., and Stayman, D. M. The role of mood in advertising effectiveness. Journal ofConsumer Research, 17, 2, 1990, 203–214.

Benassi, P. TRUSTe: an online privacy seal program. Communications of the ACM, 42,2, 1999, 56–79.

Berndt, B. Social judgment theory and the analysis of interpersonal conflict.Psychological Bulletin, 83, 6, 1976, 985–1003.

Boone, L. E., and Kurtz, D. L. Contemporary Business 2010 Update. John Wiley andSons, New York, NY, 2010.

Chaiken, S., and Maheswaran, D. Heuristic processing can bias systematicprocessing: effects of source credibility, argument ambiguity, and taskimportance on attitude judgment. Journal of Personality and Social Psychology,66, 3, 1994, 460–473.

Chase, R. B., and Dasu, S. Want to perfect your company’s service? Use behavioralscience. Harvard Business Review, 79, 6, 2001, 78–84.

Cialdini, R. B. Influence: The Psychology of Persuasion. Morrow, New York, NY, 1993.Cook, D. P., and Luo, W. The role of third party seals in building trust online. E-

Service Journal, 2, 3, 2003, 71–85.Crespo, A. H., and Bosque, I. R. E. The influence of the commercial features of the

Internet on the adoption of e-commerce by consumers. Electronic Commerce andResearch Applications, 9, 6, 2010, 562–575.

DeLone, W., and McLean, E. The DeLone and McLean model of information systemssuccess: a ten-year update. Journal of Management Information Systems, 19, 4,2003, 9–30.

Fornell, C., and Larcker, D. F. Evaluating structural equation models withunobservable and measurement error. Journal of Marketing, 18, 1, 1981, 39–50.

Hair, J. F., Black, B., Babin, B., Anderson, R. E., and Tatham, R. L. Multivariate DataAnalysis. Prentice Hall, Englewood Cliffs, NJ, 2006.

Head, M. M., and Hassanein, K. Trust in e-commerce: evaluating the impact of third-party seals. Quarterly Journal of Electronic Commerce, 3, 3, 2002, 307–325.

Heald, J. E. Social judgment theory: applications to educational decision making.Educational Administration Quarterly, 27, 3, 1991, 343–357.

Heesacker, M., Petty, R. E., and Cacioppo, J. T. Field dependence and attitude change:source credibility can alter persuasion by affecting message-relevant thinking.Journal of Personality, 51, 4, 1983, 653–666.

Heijden, V. D., Verhagen, T., and Creemers, M. Understanding online purchaseintentions: contributions from technology and trust perspectives. EuropeanJournal of Information Systems, 12, 1, 2003, 41–48.

Hovland, C. I., and Weiss, W. The influence of source credibility on communicationeffectiveness. Public Opinion Quarterly, 15, 4, 1951, 635–650.

Joo, S. H. J., Grable, J. R., and Bagwell, D. C. Credit card attitudes and behaviors ofcollege students. College Student Journal, 37, 3, 2003, 405–419.

Karahanna, E., Straub, D. W., and Chervany, N. L. Information technology adoptionacross time: a cross-sectional comparison of pre-adoption and post-adoptionbeliefs. MIS Quarterly, 23, 2, 1999, 183–213.

Kim, D., and Benbasat, I. Trust-assuring arguments in B2C e-commerce: impact ofcontent, source, and price on trust. Journal of Management Information Systems,26, 3, 2009, 175–206.

Kimery, K. M., and McCord, M. Third-party assurance: mapping the road to trust ine-retailing. Journal of Information Technology Theory and Applications, 4, 2, 2002,63–82.

Komiak, S. Y. X., and Benbasat, I. The effects of personalization and familarity ontrust and adoption of recommendation agents. MIS Quarterly, 30, 4, 2006, 941–960.

Kovar, S., Burke, K., and Kovar, B. Consumer responses to the CPA WebTrust privacyseal. Journal of Information Systems, 14, 1, 2000, 3–17.

Krosnick, J. A. Government policy and citizen passion: a study of issue publics incontemporary america. Political Behavior, 12, 1, 1990, 59–92.

Lala, V., Arnold, V., Sutton, S. G., and Guan, L. The impact of relative informationquality of e-commerce assurance seals on Internet purchasing behavior.International Journal of Accounting Information Systems, 3, 4, 2002, 237–253.

Liao, Z., and Cheung, M. T. Internet-based e-shopping and consumer attitudes: anempirical study. Information & Management, 38, 5, 2001, 299–306.

Mazis, M. B., Ahtola, O. T., and Klippel, R. E. A comparison of four multi-attributemodels in the prediction of consumer attitudes. The Journal of ConsumerResearch, 2, 1, 1975, 38–52.

Maison, D., Greenwald, A. G., and Bruin, R. H. Predictive validity of the implicitassociation test in studies of brands, consumer attitudes, and behavior. Journalof Consumer Psychology, 14, 4, 2004, 405–415.

McKnight, D. H., and Chervany, N. L. What trust means in e-commerce customerrelationships: an interdisciplinary conceptual typology. International Journal ofElectronic Commerce, 6, 2, 2001, 35–59. 2002.

MIC, Market Intelligence Centre. The trend of B2C e-commerce in Taiwan, 2010.Available at: <www.ithome.com.tw/itadm/article.php?c=58860>.

Miyazaki, A. D., and Krishnamurthy, S. Internet seals of approval: effects on onlineprivacy policies and consumer perceptions. The Journal of Consumer Affair, 36, 1,2002, 28–49.

Nöteberg, A., Christiaanse, E., and Wallage, P. Consumer trust in electronic channels.E-Service Journal, 2, 2, 2003, 46–67.

Odom, M., Kumar, A., and Saunders, L. Web assurance seals: how and why theyinfluence consumers’ decisions. Journal of Information Systems, 16, 2, 2002, 231–250.

Parra, J. F., and Ruiz, S. Consideration sets in online shopping environments: theeffects of search tool and information load. Electronic Commerce Research andApplications, 8, 5, 2009, 252–262.

Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., and Podsakoff, N. P. Common methodbiases in behavioral research: a critical review of the literature andrecommended remedies. Journal of Applied Psychology, 88, 5, 2003, 879–903.

Sachse, S. B., and Gierl, H. Can a positive mood counterbalance weak arguments inpersonal sales conversations? Journal of Retail and Consumer Services, 16, 3,2009, 190–196.

Sherif, M., and Hovland, C. I. Social Judgment: Assimilation and Contrast Effects inCommunication and Attitude Change. Yale University Press, New Haven, CT, 1980.

Sherif, C. W., Sherif, M., and Nebergall, R. E. Attitude and Attitude Change. Saunders,Philadelphia, PA, 1965.

Siero, F. W., and Doosje, B. J. Attitude change following persuasive communication:integrating social judgment theory and the elaboration likelihood model.European Journal of Social Psychology, 23, 5, 1993, 541–554.

Smith, G. F., Benson, P. G., and Curley, S. P. Belief, knowledge, and uncertainty: acognitive perspective on subjective probability. Organizational Behavior andHuman Decision Processes, 48, 2, 1991, 291–321.

TNS Market Research. The number-one sign of trust of the Internet, 2008. Availableat <www.verisign.com/static/013506.pdf>.

Toulmin, S. E. The Use of Argument. Cambridge University Press, Cambridge, UK,2003.

Wang, W. Q., and Benbasat, I. Interactive decision aids for consumer decisionmaking in e-commerce: the influence of perceived strategic restrictiveness. MISQuarterly, 33, 2, 2009, 293–320.

Williams, R., and Grimes, A. Involvement and the influence of online third-partyendorsements. International Journal of Internet Marketing and Advertising, 6, 1,2010, 65–84.

Wood, R., and Bandura, A. Social cognitive theory of organizational management.Academy of Management Review, 14, 3, 1989, 361–384.

Zhang, H. Trust-promoting seals in electronic markets: impact on online shoppingdecisions. Journal of Information Technology Theory and Applications, 6, 4, 2004,29–40.