Volunteering personal information on the internet: Effects of reputation, privacy notices, and rewards on online consumer behavior

(1)

DOI 10.1007/s11002-006-4147-1

Volunteering personal information on the internet:

Effects of reputation, privacy notices, and rewards on online consumer behavior

En Xie·Hock-Hai Teo·Wen Wan

CSpringer Science+Business Media, Inc. 2006

Abstract Accurate personal information provision is one of the most important determinants of the commercial development of the Web. However, consumers are usually reluctant to provide personal information or tend to provide false information online because of their concern about privacy violation risks. We employ a 2×2×2 experimental design to examine the effects of reputation, privacy notices, and rewards on online consumer behavior in volunteering two types of personal information on the Internet: demographic information and personally identifiable information. We discuss the theoretical and practical implications of the findings.

Keywords Consumer personal information . Privacy . Reputation . Privacy notices . Reward

1. Introduction

Accurate consumer personal information is one of the most strategic assets of a firm. With- out accurate consumer personal information, firms cannot effectively perform direct marketing, customer-relationship management, and strategic production of goods and services (Henderson and Snyder, 1999; Long et al., 1999; Milne, 1997). The importance of consumer personal information is salient as the Internet serves as a distribution channel. According to Hoffman et al. (1999), the willingness of consumers in providing personal information is one of the most important determinants of the commercial development of the Web.

E. Xie ()

Department of Marketing, School of Management, Xi’an Jiaotong University, Xi’an, Shaanxi Province, P.R. China, 710049

e-mail: [email protected] H.-H. Teo·W. Wan

School of Computing, National University of Singapore, Room 5-23, SoC 1, Building, 3 Science Drive 2. Singapore 117543

e-mail: [email protected]

(2)

However, most consumers concern about their privacy information and are reluctant to provide personal information online. According to Hoffman et al. (1999), almost 95% of Web users have declined to provide personal information to Web sites at one time or another when asked, and 40% who have provided demographic data have gone to the trouble of fabricating it. A 2001 Harris Interactive survey found that about 83% of online users have refused to give information to a business or company and even over 44% online users have avoided specific Web sites because of dubious privacy practices (Harris interactive, 2001). According to Jupiter Media Matrix’s Consumer Survey in 2002, 70% of U.S. consumers are concerned about their privacy online (Jupiter Media Matrix’s Consumer Survey, 2002). The reluctance to provide information online could be attributed to some Internet characteristics. For example, the information provided online could be combined with some information (e.g., cookies or page views behavior of individuals) that are surreptitiously collected over the Internet to profile the behaviors of individuals (Dommeyer and Gross, 2003). In addition, the collection of personal information could be performed without the consumers’ awareness or permission (Milne and Culnan, 2004). Due to these characteristics, consumers would be more concern about their privacy protection in online environment. Thus an in-depth understanding of the strategies promoting information disclosure is of paramount importance to managers as well as researchers.

In the existing literature, many researchers proposed that the “risk-benefit” tradeoff was essential to the information disclosure decision of consumers: Individuals should assess the outcomes they receive as the risk of providing personal information to firms (Culnan and Bies, 2003). Based on such an assessment, a positive net outcome should mean peoples are more likely to accept the loss of privacy that accompanies any disclosure of personal information as long as an acceptable level of risk accompanies the benefits. Implied by this

“risk-benefit” perspective of information provision is that firms or website could enhance the willingness of consumers to disclose personal information in two ways (Culnan and Bies, 2003): (1) offering attractive benefits to consumers or (2) decreasing the perceived risk of consumers in information disclosure. Thus, two types of instruments promoting information disclosure were discussed in the existing literature. From the benefit side, some researchers have suggested that firms should offer direct and immediate rewards in the form of discount coupons and bonus points to encourage consumers to register and provide personal information (Hann et al., 2003). On the other hand, several researchers and privacy advocates have proposed that firms could adopt initiatives which decrease the perceived risk of consumers in information disclosure. Based on this risk perspective, instruments such as privacy notices, privacy seals from third parties, and the Platform for Privacy Protection (P3P) are emphasized in promoting personal information disclosure (e.g., Milne and Culnan, 2004; Das et al., 2003;

Culnan and Armstrong, 1999).

These related researches provide us significant insight into online information disclosure. However, we are of the view that the effectiveness of these instruments in addressing consumers’ privacy concerns to induce them to provide accurate personal information is contingent on some contextual factors. For example, privacy concerns may be especially salient for consumers when they interact with a firm or a Web site with little or no reputation (Milne and Culnan, 2004). Our study hence draws on utility theories and reputation perspective to examine the effects of rewards, privacy notices and reputation on the online consumer behavior of volunteering personal information on the Internet. In particular, we employ a 2×2×2 experimental design to investigate how these variables affect online consumer behavior in volunteering two types of personal information: demographic information and personally identifiable information. Several preliminary understanding may be drawn from this exploratory study. First, the study demonstrates the importance of privacy notice,

(3)

rewards and reputation in promoting personal information disclosure. Second, our findings could allow both reputable firms and firms that have yet to establish their reputation to assess the relative effectiveness of privacy notices and rewards in inducing consumers to volunteer personal information, and hence, make the necessary investments in these instruments.

2. Theoretical perspectives and hypotheses

2.1. Utility-based perspective of information provision

Studies based on risk-benefit perspective or utility theories suggest consumers may relinquish some privacy in return for one-time rewards in the form of tangible payments. In several studies (e.g., Zhang and Chen, 2000), tangible rewards such as discount coupons and bonus points are identified as one of the main reasons for consumers to answer data collection questions and to engage the services of a firm. This utility perspective is called the “privacy calculus” in some privacy studies, and it states that individuals perform cost and benefit assessments of the behavior they are contemplating (e.g., providing accurate information) (Laufer and Wolfe, 1977; Lwin, 2003). Utility theories posit that people maximize their total utility when making choices, and they are willing to disclose personal information in exchange for some economic or social benefit, subject to the privacy calculus (e.g., Milne and Gordon, 1993; Culnan and Bies, 2003). In other words, individuals will exchange personal information when they perceive adequate benefits will be received in return (Culnan and Bies, 2003; Phelps et al., 2000). In line with these studies, we hypothesize that providing rewards should increase consumers’ utility, which would in turn motivate them to provide accurate information online. Hence, we hypothesize:

H1: Rewards are positively related to online users’ provision of accurate personal information.

Even with the existence of rewards, consumers would still be reluctant to disclose information if the perceive risk of the disclosure is high. Hence the instruments that decrease the perceived risk could also promote the information disclosure. From the consumer perspective, they are concern about the ways firms or websites use personal information about them (Phelps et al., 2000). Privacy notices which provide consumers with information about the firm’s information practices help to signal the commitment of firms in protecting consumer privacy and observation of fair information practice. Firms that post privacy notices at their Web sites would be deemed as unlikely to violate individuals’ privacy for fear of negative legal consequences or sanctions by third-party authentication services like TRUSTe. Hence, privacy notices help engender consumers’ trust to reduce the risk of information disclosure (Milne and Culnan, 2004). In light of this, we believe that privacy notices would have a positive effect on individuals’ information provision. Some theoretical and empirical studies have found privacy notices useful in alleviating consumers’ privacy concerns (e.g., Milne and Culnan, 2004; Phelps et al., 2000; Hoffman et al., 1999). Hence, we hypothesize:

H2: Availability of privacy notices are positively related to online users’ provision of accurate personal information.

2.2. Reputation perspective of information provision

In the marketing literature, reputation is defined as the extent to which firms and people in the industry believe a firm is honest and concerned about its customers (Doney and Cannon,

(4)

1997). Reputation has been found to be the key factor for engendering trust in firms among consumers in both the traditional marketing exchange (Doney and Cannon, 1997; Ganesan, 1994) and electronic commerce contexts (Jarvenpaa and Tractinsky, 1999). Mitchell and Vincent-Wayne (1999) suggest that reputation affects cognitive perceptions of quality and in turn engenders trust. Moorman et al. (1992) make similar comments about reputation being an indicator of reliability.

Owing to the lean nature of the electronic environment relative to the traditional face- to-face market, a high level of trust is more important in electronic commerce (e.g., Roboff and Charles, 1998; Milne and Boza, 1999). Creating trust has been identified as an effective approach for organizations to increase information disclosures from consumers and obtain the base of successful online strategies (Milne and Culnan, 2004; Hoffman et al., 1999; Sultan and Moraaj, 2001). Because reputation is an important property of online firms to engender trust, we expect high reputation to be related to more effective online information solicitation.

A firm’s reputation is a strategic asset that takes time to build and requires significant investment. It is therefore unlikely that a reputable firm would jeopardize its reputation by behaving opportunistically. Hence, from this standpoint, high reputation serves as a means to reduce uncertainty and generate a feeling of trust among consumers to engage in transactions with the firm. The link between reputation and transactional activities is supported by empirical evidence from the field of industrial channel dyads (Ganesan, 1994; Anderson and Weitz, 1989) and the electronic commerce environment (Bolton et al., 2004; Stewart, 2003). Based on these observations, we hypothesize:

H3: The reputation of a firm or Web site is positively related to online users’ provision of accurate personal information.

Firms which do not have a high reputation would normally employ other instruments to entice consumers to engage in transactional activities with them. In the electronic commerce world, several online firms have used the lure of rewards and the peace of mind promised by privacy notices to attract consumers to provide information online. To the extent that these instruments are effective, we believe that they should have a more significant effect on online information provision by consumers for firms of low reputation than for firms of high reputation. Consumers may rely on the reputation of the firm as a signal that provides assurances their information is safe (Gefen et al., 2003; Shapiro, 1987). High reputation could engender trust in consumers about a firm’s services and practices, thus lessening the need to invest in other instruments. Milne and Culnan (2004) also argue that reputation is a substitute for reading privacy notices. Hence, we hypothesize:

H3a: The effects of rewards on online users’ provision of accurate personal information are stronger for firms of low reputation than for firms of high reputation.

H3b: The effects of privacy notices on online users’ provision of accurate personal information are stronger for firms of low reputation than for firms of high reputation.

3. Method

We used a 2×2×2 factorial design to test our hypotheses. Table 1 depicts the eight treatment combinations as well as the number of cases of each combination collected from our experiment.

(5)

Table 1 Experiment design

Treatment Privacy notice No privacy notice

High Reward 20 17

Reputation No reward 20 17

Low Reward 17 20

Reputation No reward 18 18

3.1. Operationalization of variables

We operationalized the independent variables using the vignette technique, which uses short scenarios in written or pictorial form to elicit perceptions, opinions, beliefs and attitudes to typical situations (Hill, 1997). Vignettes are particularly useful for eliciting actions or intentions for different situational contexts; clarifying individual judgment, especially in relation to moral dilemmas; and discussing sensitive experiences in comparison to the norms of the day (Finch, 1987). With the use of vignettes, respondents can easily express their own perceptions on topics very familiar to them while remaining detached. The advantage of this technique is that respondents do not have to bias their responses or feel compelled to give socially approved answers since they do not perceive any danger of devaluing their personal image by giving sincere answers (Alexander and Becker, 1978). In our experiment, we aimed to elicit online consumers’ willingness to provide accurate information when confronted with the need to provide information before they could use the services of an online store. As improperly written or designed vignettes might allow subjects to guess the experiment intention, we took special care to write and test the vignettes in a pilot study involving 20 students before conducting the actual study. All but one was unable to guess our hypotheses correctly.

In our hypotheses, there are three hypothesized determinants of online information provision: Reward, Reputation and Privacy Notices. Notably, all three independent variables are categorical variables. In our experiment, we set Reward to one if the firm offered a $20 gift voucher for consumers who agreed to provide accurate information, and zero otherwise.

Likewise, Privacy Notices equaled one if the firm had a secure connection, possessed a privacy statement, and had obtained third-party authentication of its privacy statement by TRUSTe.

Reputation equaled one if the firm had been established for over 50 years, was a member of the Conglomerate of 200 Shopping Malls,¹and was dedicated to customer satisfaction.

Reputation equaled zero if the firm had been established for about five years, its Web site only offered delivery in one country, and it made no mention of customer satisfaction. Table 2 presents the operationalization of the independent variables in detail.

Each treatment was illustrated by a different vignette presented via an Internet-based system developed using Active Server Pages (ASP). For high settings in Reputation, Reward and Privacy Notices, the Internet-based system presented to subjects a vignette that described a firm with the following characteristics: established for over 50 years, listed as a member of the Conglomerate of 200 Shopping Malls, and dedicated to customer satisfaction (Reputation);

offered a $20 gift voucher to consumers who agreed to provide accurate information (Reward);

implemented a secure connection, had a privacy statement, and had obtained third-party authentication of its privacy statement by TRUSTe (Privacy Notices).

1“Conglomerate of 200 Shopping Malls” is a list of 200 major shopping malls in Singapore. Being included in the list is an indication of a firm’s good reputation.

(6)

Table 2 Operationalization of independent variables

Company information in the internet-based experiment System

Reputation

High Established for over 50 years

(Reputation=1) Member of the conglomerate of 200 shopping malls Dedicated to customer satisfaction

Low Established for about five years Web site only delivers to one country (Reputation=0) No mention of customer satisfaction Privacy Notice

Present Secure connection to server (Privacy Notice=1) Presence of a privacy policy

TRUSTe certification Absent No secure connection to server (Privacy Notice=0) No privacy policy

No TRUSTe certification Reward

Present $20 gift voucher offered (Reward=1)

Absent No gift voucher offered (Reward=0)

The dependent variable in the study is the willingness of online consumers to provide personal information. Phelps et al. (2000) suggest personally identifiable information and demographic information are two primary categories of personal information which would cause privacy concern and usually, personally identifiable information would relate more closely to privacy concern of consumers (Nowak and Phelps, 1992). Though these two categories may vary in the degree to which each evoke privacy concerns, marketers do need to assimilate both personally identifiable information which would be useful to provide one-to-one service and demographic information which has a more direct relationship to message and media strategy (Phelps et al., 2000). Thus we included the willingness to provide personally identifiable information and demographic information as dependent variables in the study. The dependent variables were operationalized with the following procedure: (1) after reading the vignette information, the subjects were shown a list of several types of information including demographic information (e.g., Hobbies, Books Read, Possessions, and Holiday Destinations; Postal Code, etc.) and personally identifiable information (e.g., Credit Card Number, Office Phone, Office Fax, etc.). A full list of the types of information is given in the appendix; it includes 14 types of demographics and 11 types of personally identifiable information. Every item in the appendix represents one type of information. (2) For every type of information (such as Name, Postal Code, or Hobbies), the subjects were asked whether they wanted to provide accurate information or not. The answer would be

“Provide” or “Not Provide”. (3) For every subject, we counted how many types of information that he/she was willing to provide. By computing the ratio of the number of items on which the subject was willing to provide accurate information to the total number of items, we could get the willingness of the subject to provide accurate information in the respective categories. To elaborate, the first dependent variable, i.e., willingness to provide demographic information, was computed as the ratio of the number of items on which the subject was willing to provide

(7)

accurate information to the total number of demographic information items that were asked of the subject. The second dependent variable, i.e., willingness to provide personally identifiable information, was similarly computed.

The categorization of the items of demographic and personally identifiable information in the study was basically based on categories used in the Federal Trade Commission (1998) survey. Notably, Phelps et al. (2000) suggest an important characteristic of personally identifiable information is the perceived sensitivity of the information would be higher than that of demographic information. Milne (1997) further contends that the relative sensitivity of the information requested could impact the willingness of information disclosure. Because there would be different understandings about the sensitivity of the given type of personal information between U.S. and Singapore, it is hence necessary to check the validity of the categorization by comparing the perceived sensitivities of items for personally identifiable and demographic information. Thus we have asked the subjects to indicate their understanding on the sensitivities of each information item in the experiment. The perceived sensitivities were anchored on a seven-point Likert scale, ranging from not sensitive at all (1) to very sensitive (7).

The ASP used in our Internet-based system allowed program logic and scripting logic to be included in standard HTML pages. These capabilities allowed us to generate the vignettes randomly to the subjects, ensure that each subject had viewed the treatment conditions before they were allowed to proceed, and ensure that the subjects had answered all the questions before leaving the experiment. The features also allowed us to be certain that the subjects had read the vignette completely before they gave their responses to the questions on whether they would provide accurate information or not.

3.2. Subjects

The subjects were recruited from the customer database of a professional special interest Web site (miexchange.net) in Singapore. An electronic mail stating the purpose of our study was sent to 400 potential subjects to invite them to participate in our experiment. We omitted those responses that failed to meet our criteria of (1) they must be greater than 13 years old, and (2) they must be Singaporean or Singapore permanent residents. We selected subjects on the above criteria as they would be more likely to comprehend our instructions. Finally, a total of 147 subjects were chosen from 400 potential subjects. The selected subjects were informed of their eligibility and given our Internet-based system website address, instructions, and password to access our system. Subjects were also told that they would be awarded a $15 in compensation for their time and effort in participating in the study. Our Internet-based system generated the firm vignette information randomly so that each respondent had an equal and independent chance of being put into any of the eight scenarios. Upon completion of the survey questions, a cashier order of $15 was sent to the address provided by the subject.

To capture the background of the subjects, we have included demographic questions regarding subjects’ gender, age, Internet usage. The age of the respondents ranges from 13 to 48 years, with the average age being 25.7 years. 87 were males and 60 were females. All the respondents accessed the Internet at least once a day, and thus were familiar with Internet usage (5.32 on a Likert scale of 1 to 7). Of the 147 respondents, 46 had purchased on the Internet before.

Some respondents of the research are quite young (above 13). We included the younger subjects in the research with consideration that the average age of web-user in Asian is relatively low. Recently more and more very young Asian people have experience in Web surfing. According to the survey by CNNIC at the start of 2004, about 18.8% Web users are under 18 in China. In Taiwan, about 52% web users are 13 to 29 years old. The characteristic

(8)

has been noticed by researchers in related fields. At the surveys mentioned in Hoffman et al.

(1999), the age of respondents is above 16. At the research of Hann et al. (2003), the average age of Singaporean respondents is 23.1. In order to reflect the true profile of the web-users in Singapore, our research needs to include younger respondents. According to the U.S. Child Online Privacy Protection Act, 13 is the minimum age from which opinions can be solicited.

Thus we set 13 as the lowest age of the subjects in the research.

3.3. Experiment procedure

After logging into our Internet-based system, the eligible subjects were asked to read the instructions and the descriptions in the vignette carefully. The Internet-based experiment system would log the activities of the subject at the Web site. For example, if the subjects had not read the privacy statement, they would not be able to complete the questionnaire.

After having read all the descriptions in the vignette, the subjects were then asked to indicate their decision to provide accurate information for every item on which users would be asked to provide information online. The subjects were also asked to indicate how sensitive they perceived the information relating to each item was.

4. Results

4.1. Manipulation checks

Several questions were asked in the questionnaire for the purpose of manipulation checks.

Results show that all the treatments were manipulated correctly. Subjects in the high reputation treatment perceived the firm to be more reliable, more trustworthy, and more financially sound than the subjects in the low reputation treatment did (t-statistic=11.034,p<0.01).

Subjects in the high reward treatment found the incentive to provide information to be higher than subjects in the no reward treatment did (t-statistic=6.596,p<0.01). Similarly, subjects in the privacy notice treatment believed the firm was less likely to violate their privacy and could protect their data better than the subjects (scale reversed) in the no privacy notice treatment did (t-statistic= −5.419, p<0.01). In order to test the validity of our categorization of personal information, we ask the subjects to rate the perceived sensitivity of each type of information on a seven-point Likert scale. According to Phelps et al. (2000), the perceived sensitivity of personally identifiable information would be higher than that of demographic information. We did find that average perceived information sensitivity of demographic information is significant lower than that of personally identifiable information. The result indicates the validity of our categorization (t-statistic=16.754,p<0.01).

4.2. Multivariate regression model

Multivariate regression was performed on both the demographic information sample and the personally identifiable information sample. A 5% significance level was used for all statistical tests. The regression model was as follows:

Action=β0+β1Rew+β2P N+β3Rep+β4Rep^∗Rew+β5Rep^∗P N

(9)

Table 3 Descriptive statistics of demographic information sample

Privacy notice No privacy notice

Treatment Mean Std. dev Mean Std. dev

High Reward Action 0.8770 0.0910 Action 0.6905 0.2272

Reputation No Reward Action 0.9370 0.0709 Action 0.7976 0.1593

Low Reward Action 0.8496 0.1812 Action 0.6325 0.1954

Table 4 Descriptive statistics of personal identification information sample Privacy notice No privacy notice

Treatment Mean Std. Dev Mean Std. dev

High Reward Action 0.6364 0.1208 Action 0.4697 0.0643

Low Reward Action 0.4091 0.1782 Action 0.3262 0.1930

Action denotes a subject’s overall choice in providing accurate personal information to the questions that were asked of him/her (overall choice means the ratio of the number of items on which the subject was willing to provide accurate information to the total number of items). Rew, PN and Rep denote Reward, Privacy Notice and Reputation respectively. As mentioned earlier, all three are categorical variables.

Tables 3 and 4 show the descriptive statistics of Action in different treatment cells.

In the demographic information sample and the personally identifiable information sample, on average, the proportion of questions that elicited accurate information provision was higher when Reward, or Reputation or Privacy Notice was high. The difference between Action in treatment cells with and without the three conditions was more evident when the subjects were requested to provide personally identifiable information. In addition, Table 3 and 4 show us that in every treatment, subjects are more willing to disclose demographic information rather than personally identifiable information. This is consistent with the argument of Phelps et al. (2000) which suggests consumers would be more willing to disclose demographic information than personally identifiable information because there typically would be lower privacy concern associated with disclosure of demographic information.

Table 5 shows the regression estimation results and Table 6 summarizes the hypotheses test results. The overall regression model is statistically significant for the two samples. The independent variables account for a sizeable 46.30% and 37.40% of the variance in accurate personal information provision in the personally identifiable information sample and the demographic information sample respectively. An assessment of the variance inflation factor values for all independent variables showed that none of the values was greater than 3. We found no support for the existence of multi-collinearity.

(10)

Table 5 Estimation results for personal identification information sample and demographic information sample

Action on personal

identification Action on demographic information questions information questions

Variable Coef T -value Coef T -value

Reward 0.181^∗∗ 4.11 −0.057 −1.45

Privacy notice 0.128^∗∗ 2.90 0.228^∗∗ 6.55

Reputation 0.190^∗∗ 3.41 0.117^∗∗ 2.75

Reputation^∗Reward −0.136^∗∗ −2.55 0.014 0.28

Reputation^∗Privacy notice 0.158^∗∗ 2.61 −0.066 −1.36

R-square 46.30% 37.40%

∗∗Significant at 5%-level

Table 6 Summary of hypotheses test results

Personal identification Demographic

Hypothesis information information

H1: Reward->Action Supported Not supported

H2: Privacy Notice->Action Supported Supported

H3: Reputation->Action Supported Supported

H3a: Reputation^∗Reward->Action Supported Not supported H3b: Reputation^∗Privacy Notice->Action Not supported Not supported

As hypothesized in H1, we found rewards positively related to an online user’s provision of accurate personal information. Our test results in Table 5 provide limited support for H1:

the positive relation between rewards and provision of personally identifiable information is significant (Coef.=0.181,p<0.05). However the results do not support a significant positive relation between rewards and provision of demographic information (Coef.= −0.057, p

>0.05). The results support H2 and H3: the adoption of privacy notices is positively related to the provision of personally identifiable information (Coef.=0.128, p<0.05) and that of demographic information (Coef.=0.228, p<0.05). Further, reputation is positively related to the provision of personally identifiable information (Coef.=0.190, p<0.05) and demographic information (Coef.=0.117, p<0.05).

The results in Table 5 show limited support for H3a: as expected, reputation has a negative moderating influence on the effects of rewards on provision of accurate personally identifiable information, and no moderating effects on demographic information provision.

The results in Table 5 do not support H3b. Interestingly, contrary to our hypothesis, the results show reputation having a positive moderating influence on the effects of privacy notices on accurate personally identifiable information provision. The results suggest privacy notices in our sample have a greater positive impact on accurate personally identifiable information provision for firms of high reputation than it does for firms of low reputation.

(11)

5. Discussions, implications and limitations

We have sought to examine the factors affecting the consumer behavior of providing personal information online from the utility and reputation perspectives. Findings from our experiment reveal that rewards, privacy notices, and reputation greatly influence consumers’ intention to provide accurate personal information over the Internet, and such effects vary according to the sensitivity of the requested information.

Rewards, in the form of a monetary voucher, were found to have a positive impact on online consumers’ decision to provide accurate personal information for personally identifiable data but not for demographic data. In line with utility-based studies, it appears that consumers are willing to risk privacy invasion in revealing their personally identifiable information in exchange for tangible rewards. For demographic information, the ratio of accurate information the subjects were willing to provide in relation to the total number of questions was quite high (ranges from 0.76 to 0.94) in treatment cells with no rewards in our experiment. One plausible explanation is that since demographic information is not specific to an individual, possession of such general personal information by firms does not equip them with the ability to trace or contact the individual, and thus the risk of privacy invasion is quite low. Hence, online consumers might not resent providing such information—even without rewards—in exchange for better services. The findings suggest that the effectiveness of rewards in encouraging consumers to provide personal information is contingent upon the type of information which online stores intend to solicit.

Privacy notices significantly boost consumers’ decision to provide accurate responses in both demographic information and personally identifiable information. Consistent with other studies (e.g., Milne and Gordon, 1993; Phelps et al., 2000; Milne and Culnan, 2004), our results indicate that adoption of privacy notices is a basic pre-requisite for firms interested in soliciting information from their online consumers. Regardless of the sensitivity of information, privacy notices are instrumental to the solicitation of accurate information from online consumers.

Our study also reveals a highly positive relationship between the reputation of a company and online consumers’ decision to reveal accurately their personal information. This finding is in line with the argument of Gefen et al. (2003) that reputation of the vendor would breed trust and in turn, facilitate information provision. To be effective in the electronic marketplace in general, and information solicitation in particular, an online firm should judiciously build and guard its reputation over time through consistent honest behavior or affiliation with more established firms. Again, our study shows that online consumers are willing to provide accurate demographic information regardless of the reputation of the online firm.

Our findings on the moderating influence of reputation on the effects of rewards and privacy notices on accurate information provision are counter-intuitive and interesting. As expected, rewards play a more important role for firms of low repute than for firms of high repute in the solicitation of accurate personally identifiable information. This finding suggests that new online firms that have yet to establish their reputation should offer some monetary incentives to induce or attract online consumers to provide accurate information about themselves.

Interestingly, privacy notices play a more significant role for firms of high repute than for firms of low repute when soliciting accurate personally identifiable information. This

(12)

counter-intuitive finding could be attributed to consumers’ awareness of privacy violations by online reputable firms such as Microsoft, Doubleclick.com, and Yahoo.com, all of which attracted extensive media coverage. Online consumers may also believe that more established firms might have more capabilities to track and analyze their online behavior compared to less established ones. This finding suggests that reputable firms should not be complacent and should do more to assure online consumers that the information they provide would not be misused or abused.

While our study makes use of real Internet users, care and caution should still be exer- cised when interpreting the results because privacy concerns and attitudes toward providing information online may vary according to the political, economic, legal and cultural contexts.

Because Singapore has a vision of being an international electronic commerce hub, thus the government has taken great efforts to build up a legal and policy framework that creates an online environment of trust, predictability and privacy protection in Singapore. The Electronic Transactions Act (ETA) was enacted on 10 July 1998 to create the legal framework for electronic commerce transactions. Now Singapore has been the front runner in protecting online privacy through legislation in Asia (Gomez, 2004). Additionally, Singaporean consumers express higher privacy concern in online environment. According to a study of Xia et al. (2003), Singaporean consumers rank security and privacy as the two most important determinants of their online shopping decision. Further, many web users in Singapore are teenage and young adults (Xia et al., 2003). Thus as many empirical studies conducted in Singapore (e.g.

Hann et al., 2003; Xia et al., 2003 etc.), the subject of the study is also relatively young. All these legal, social, and web-users’ characteristics in Singapore would exert influence on the testing results. Where possible, this study should be replicated in other countries to ascertain the impact of privacy initiatives and rewards on online consumers’ information provision behavior.

While our study sheds some light on how firms and Web sites could facilitate consumers to providing personal information, it has certain limitations. In particular, we have not explored the impact of the relevance of the requested information to the transaction at hand on the willingness of consumers to disclose personal information. As we could imagine, if the requested information were relevant to the transaction and could help consumers transact in a time-saving and convenient manner, consumers would be willing to disclose personal information (Hann et al., 2003). It would be interesting to investigate this aspect in future research.

6. Conclusions

Accurate information provision is a pillar of electronic commerce success. This study provides strong evidence on the effects of rewards, privacy notices, and reputation on the provision of accurate personally identifiable information. Our study is novel to the extent that few studies have examined the issue of providing accurate personally identifiable and demographic information. Moreover, very few studies have examined the relative effectiveness of the reward and privacy notice instruments for different firm types. Our findings suggest that online firms should employ different strategies to solicit information according to the nature of the firm.

(13)

Appendix: Mean perceived sensitivity of information requested

Mean of Perceived Information Item Type of Information Sensitivity^∗ Credit Card Number Identifier (Financial) 6.5238

Home Phone Number Identifier 6.2925

Address Identifier 5.9796

Mobile Identifier 5.9116

Office Fax Identifier 5.5850

Credit Card Expiry Date Identifier (Financial) 5.5646

Office Phone Identifier 5.5646

Personal Fax Number Identifier 5.1429

Home Page Address Identifier 4.3809

Email Identifier 4.2517

Name Identifier 3.1973

Income Demographics (Financial) 4.6939

Postal Code Demographics 3.9456

Date of Birth Demographics 3.6735

Job Title Demographics 3.1429

Number of Children Demographics 3.1292

Occupation Demographics 3.0408

Country of Residence Demographics 2.9864

Level of Education Demographics 2.8299

Marital Status Demographics 2.7619

Possessions Demographics (Preference) 2.2993 Holiday Destinations Demographics (Preference) 2.2789 Books Read Demographics (Preference) 2.2449

Hobbies Demographics (Preference) 2.0476

Gender Demographics 1.7823

∗1-Not sensitive at all, 7-Very sensitive to me.

References

Alexander, C. S., & Becker, H. J. (1978). The use of vignettes in survey research.Public Opinion Quarterly, 42, 93–104.

Anderson, E., & Weitz, B. (1989). Determinants of continuity in conventional industrial channel dydads.

Marketing Science, 8, 310–323.

Bolton, G. E., Katok, E., & Ockenfels, A. (2004). How effective are electronic reputation mechanisms? An experimental investigation.Management Science, 50, 1587–1602.

Culnan, M. J., & Armstrong, P. K. (1999). Information privacy concerns, procedural fairness and impersonal trust: An empirical investigation.Organization Science, 10, 104–115.

Culnan M. J., & Robert J. B. (2003). Consumer privacy: Balancing economic & justice considerations.Journal of Social Issues, 59(2), 323–342.

Das S. et al. (2003). The effect of interpersonal trust, need for cognition, and social loneliness on shopping, information seeking and surfing on the web.Marketing Letters, 14, 185–202.

(14)

Dommeyer, C. J., & Gross, B. (2003). Consumer knowledge, awareness, and use of privacy protection strategies.Journal of Interactive Marketing, 17(2), 34–51

Doney, P. M., & Cannon, J. P. (1997). An examination of the nature of trust in buyer-seller relationship.Journal of Marketing, 61, 35–51.

Finch, J. (1987). The vignette technique in survey research.Sociology, 21, 105–114.

Ganesan, S. (1994). Determinants of long-term orientation in buyer-seller relationships.Journal of Marketing, 58, 1–19.

Gefen, D., Karahanna, E., & Straub, D. W. (2003). Trust and TAM in online shopping: An integrated model.

MIS Quarterly, 27(1), 51–90

James G. (2004). Dumbing down democracy: Trends in internet regulation, surveillance and control in asia, Paper presented at the Centre for Southeast Asian Studies Seminar series, Monash Asia, Institute, Monash University on 11 March 2004.

Harris Interactive (2001). Consumer privacy attitudes and behaviorsfor the privacy leadership initiative (PLI).

Harris Interactive.

Henderson, S. C., & Snyder, C. A. (1999). Personal information privacy: Implications for MIS managers.

Information & Management, 36, 213–220.

Hill, M. (1997). Research review: Participatory research with children.Child and Family Social Work, 2, 171–183.

Hoffman, D. L., Novak, T. P., & Peralta, M. (1999). Building consumer trust online.Communications of the ACM, 42(4), 80–85.

Hann, I. H., Hui, K.-L., Lee, T. S., Png, I. P. L. (2003). The value of online information privacy: An empirical investigation. Working Paper at WUSTL.

Jarvenpaa, S. L., & Tractinsky, N. (1999). Consumer trust in an internet store: A cross-cultural validation.

Journal of Computer-Mediated Communication, 5.

Jupiter Media Matrix. (2002) March Consumer Privacy Survey. Retrieved March 10. 2003from Http:// re- tailindustry.about.com/library/b1/02q2/b1 jmm060302.htm

Laufer, R. S., & Wolfe, M. (1977). Privacy as a concept and a social issue: A multidimensional developmental theory.Journal of Social Issues, 33, 22–42.

Long, G., Hogg, M. K., Hartley, M., & Angold, S., (1999) Relationship marketing and privacy. Exploring the thresholds, Journal of Marketing Science,Applied Marketing Science, 5, 4–20.

Lwin M. O., & Williams, J. D. (2003). A model integrating the multidimensional developmental theory of privacy and theory of planned behavior to examine fabrication of information online.Marketing Letters, 14, 257–272

Milne, G. R. & Gordon, M. E. (1993). Direct mail privacy-efficiency trade-offs within an implied social contract framework.Journal of Public Policy & Marketing, 12, 206–215

Milne, G. R. (1997). Consumer participation in mailing lists: A field experiment,Journal of Public Policy and Marketing, 16, 298–309.

Milne, G. R., & Boza, M. (1999). Trust and concern in consumers’ perceptions of marketing information management practices.Journal of Interactive Marketing, 13, 5–24

Milne, G. R., & Culnan, M. J. (2004). Strategies for reducing online privacy risks: Why consumers read (or don’t read) online privacy notices.Journal of Interactive Marketing, 18(3), 15–29.

Mitchell & Vincent-Wayne. (1999). Consumer perceived risk: Conceptualisations and models.European Journal of Marketing, 33, 163–195.

Moorman, Z. G. & Deshpand´e, R. (1992). Relationships between providers and users of market research: The dynamics of trust within and between organizations.Journal of Marketing Research, 29, 314–328.

Nowak, G. J., & Phelps, J. (1992). Understanding privacy concerns.Journal of Direct Marketing, 6(4), 28–39.

Phelps, J., Nowak, G., & Ferrell, E. (2000). Privacy concerns and consumer willingness to provide personal information.Journal of Public Policy & Marketing, 27–41.

Roboff, G., & Charles, C. (1998). Privacy of financial information in cyberspace: Banks addressing what consumers want.Journal of Retail Banking Services, 20, 51–56.

Shapiro, S. P. (1987). The social control of impersonal trust.American Journal of Sociology, 93(3), 623–658.

Stewart, K. J. (2003). Trust transfer on the world wide web.Organization Science, 14(1), 5–17.

Sultan, F., & Mooraj, H. A. (2001). Designing a trust-based e-business strategy,Marketing Management, 10(4), 40–46.

Y. Xia et al. (2003). Consumer preference for commercial web site design: An Asia-pacific perspective.Journal of Consumer Marketing, 20(1), 10–27.

Zhang, Y., Wang, C. L., & Chen, J. (2000). Consumers’ responses to web-based data collection efforts and factors influencing the responses.Journal of International Marketing and Market Research. 25, 115–123