Understanding the Intention of Information Contribution to Online ...

5 downloads 26109 Views 199KB Size Report
Understanding the Intention of Information Contribution to Online Feedback. Systems from ... National University of Singapore ... identify benefit and cost factors influencing consumer ...... Information Technology in Small Business: Theory and.
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

Understanding the Intention of Information Contribution to Online Feedback Systems from Social Exchange and Motivation Crowding Perspectives Yu Tong, Xinwei Wang, Hock-Hai Teo Department of Information Systems National University of Singapore {tongyu, wangxw, teohh}@comp.nus.edu.sg

Abstract The online feedback system (OFS) has been touted to be an effective artifact for electronic word-of-mouth (EWOM). Accumulating sufficient detailed consumption information in the OFS is essential to the success of OFS. Yet, past research has focused on the effects of OFS on building trust and promoting sales and little knowledge about information provision to OFS has been developed. This study attempts to fill this gap by developing and testing a theoretical model to identify the possible antecedents that lead to the intention of consumers’ information contribution to OFS. The model employs social exchange theory to identify benefit and cost factors influencing consumer intention, and motivation crowding theory to explore the moderating effects from environmental interventions that are embodied in OFS. Our preliminary results in general provide empirical support for the model. Practical implications are offered to OFS designers for system customization.

1. Introduction Word-of-mouth (WOM) communication, a form of consumer-to-consumer information exchange, influences product evaluations and purchase decisions [27]. In the past decade, the Internet has very much extended the boundary of WOM impact. Instead of neighbors and friends, people can share their opinions and experiences with complete strangers who are socially and geographically distant [33]. Research shows that electronic word-of-mouth (EWOM) has become an important factor shaping consumer purchase behavior [6, 22]. The online feedback system (OFS) is defined as the mechanism which utilizes “the Internet’s bidirectional communication capability to artificially engineer largescale, word-of-mouth networks in which individuals

share opinions and experiences on a wide range of topics, including companies, products and services” [10, p.1407]. The OFS has been touted as one of the most powerful platforms on which EWOM takes place [13]. Some noteworthy OFS examples include eBay, Amazon, and ePinions. Different OFSs offer a variety of settings to solicit user’s feedback, such as detailed reviews, short comments and ratings. In this study, we focus on the type of OFS that allows users to provide detailed product reviews on almost every aspect of consumption, because such reviews contain much richer information than simple comments or ratings do. Sufficient high-quality review repository is essential to the success of OFS. Research shows that the value of OFS lies in the product information from consumers that is capable of revealing the quality of products and services [46]. Different from traditional face-to-face WOM which is embedded in preexisting social relationship, OFSs are sustainable only through individual’s volitional communication since no one can force others to provide reviews [40]. Whereas extant research has focused considerable attention on the impact of OFSs on building trust and promoting sales [3, 4, 6, 22], little emphasis has been placed on the antecedents of detailed consumption information provision in OFSs [25]. Among a few exceptional prior studies which attempt to understand the general WOM behavior [12, 25, 42], focus is only placed on the motivators for consumers participating in WOM communication. However, research in knowledge sharing suggests that costs are important antecedents in determining information contribution [29]. Attempting to fill the above gaps, our study draws on social exchange theory and motivation crowding theory to formulate a theoretical model to explore both costs (demotivators) and benefits (motivators) of consumers’ information contribution to OFS and the conditions under which these factors are significant. We empirically tested the model with an experiment on 176 students in a large university. This study makes

1530-1605/07 $20.00 © 2007 IEEE

1

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

contributions in both academic and practical areas. Academically, this study advances theoretical development in the area of EWOM and OFSs in particular by explicating the benefits and cost factors antecedent to consumers’ participations in EWOM and by contributing a set of reliable and valid instruments. For practitioners, the results of this study highlight some system design guidelines that could help system designers customize the OFS in order to generate more WOM.

2. Literature review This study examines a consumer’s intention to provide detailed consumption information to OFS. Behavior intention is defined as “the strength of conscious plans to perform the target behavior” [24, p.176]. In general, when an individual has stronger intention to engage in a behavior, he/she is more likely to perform it [1]. In mainstream information system research, intention has been well established as a predictor of actual usage [44]. In this study, we focus on understanding different factors affecting consumer’s intention to contribute product reviews to OFSs from the perspectives of social exchange theory and motivation crowding theory.

2.1 Social exchange theory and OFSs Social exchange theory (SET), one of the major theoretical perspectives in psychology, is used to understand individual behaviors in the process of resource exchange [8]. According to SET, individual behaviors are guided by a relative simple principle: increase outcomes they perceive positive (benefits) and decreases outcomes they perceive negative (costs) [30]. Research shows that consumer WOM can be regarded as a social exchange behavior whereby consumer provides their consumption experience in exchange of some valuable outcomes (such as status) [18]. OFSs have become an eminent online platform to facilitate many-to-many information exchange [32]. Applying SET, we expect consumers are motivated to provide product reviews if benefits obtained exceed costs associated with the contribution [40]. 2.1.1 Benefit Factors. Benefits in SET, defined as exchange outcomes which provide positive values [30] tend to act as motivators for people to participate in social exchange [8, 28]. According to motivation literature, motivators can be classified into three categories: intrinsic, internalized extrinsic and extrinsic motivation [9, 37].

When intrinsically motivated, people may perform an activity for their own sake and derive pleasure and satisfaction inherent in the activity [45]. In the context of OFS, feedback contributors may perceive satisfaction by being able to help other consumers to make a better purchase decision or influence the company [25, 42]. However, people who are extrinsically motivated perform tasks as a means to other ends. They focus on outcomes that may be obtained as a result of the behavior rather than engaging in it for its inherent satisfaction [45]. In the behavior literature, extrinsic motivation is characterized by a strong focus on reward contingencies, such as pecuniary compensation [38]. As OFSs have increasingly recognized the value of EWOM and implemented economic rewarding mechanisms to promote EWOM, consumers may choose to write reviews for monetary benefits. Different from strictly extrinsic motivation, internalized extrinsic motivation is treated as a combination of intrinsic and extrinsic motivation. An individual, who is driven by this type of motivation, experiences an internalization process that assimilates and transforms external incentives into his or her own motives [9]. According to Roberts et al. [37], the internalization process can take place when an individual strives to earn external recognition from others. In OFSs, information contributors have the opportunity to gain status and acceptance among other people through certain OFS artifacts that indicate their expertise and therefore may experience an internalization process [25]. 2.1.2 Cost Factors. Cost in SET is defined as negative outcomes from exchange behavior and hence decreases the frequency of the behavior [30]. Specifically, costs can be in the form of intrinsic cost and opportunity cost [30]. Intrinsic cost, such as fatigue, pain, and unpleasantness, is intrinsic to the performance of the exchange behavior. In order to contribute product feedback, a consumer needs to cognitively retrieve her knowledge and evaluation of a particular product from the memory. This cognitive process is extensive and may cause negative psychological effect such as annoyance and irritation. Opportunity cost is the loss of benefit that individual can get from engaging in other behaviors [26]. Providing review in OFSs may demand substantial time and effort for consumers to materialize or codify their cognitive expertise and knowledge into the online system. This transformation process could deprive other benefits from alternative behaviors [28] and hence constitutes a type of opportunity cost.

2

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

2.2 Motivation crowding theory and OFS Motivation crowding theory (MCT) suggests that under different conditions, external interventions that originate from outside a person may undermine or strengthen intrinsic motivation [19, 20]. These two opposite effects towards people’s intrinsic motivation are defined as: crowding-out effect and crowding-in effect [20]. Crowding-out effect occurs when people perceive they are controlled by external conditions, their self-determination and/or self-esteem are undermined and consequently, their intrinsic motivation to perform the task is reduced. The opposite, crowding-in effect, occurs when people feel they are supported by external interventions. In such a case, their self-determination and/or self-esteem are appreciated as they are given more freedom and respect to perform the task. Therefore, people’s intrinsic motivation to the task increases [19]. It is acknowledged that an external intervention such as similar contribution from other people can affect contributors’ intrinsic motivation (for example, altruism) and, in turn, their propensity to engage in the behavior [2]. MCT has been applied to online environment recently. Dellarocas and Narayan [11] suggest that external means such as availability of information from other sources can decrease people’s intrinsic motivation to post online movie reviews. Therefore, we expect that MCT construct can moderate the relationships between certain SET constructs and the intention to contribute product reviews in OFSs. In this study, we employ the number of existing product reviews as the external intervention and study its moderating impact on the relationship between intrinsic benefits and intention to contribute in OFSs. The number of existing product reviews has been viewed as a unique characteristic in OFSs [11]. Different from offline WOM, most OFSs aggregate all the reviews for the same product. Therefore, before posting a new review for a product, a consumer is inevitably cognizant of the number of existing reviews. In this case, consumer’s self-esteem and consequently, intrinsic motivation, may be affected by this external intervention from OFSs.

3. Research model and hypotheses The research model, depicted in Figure 1, is developed on SET and MCT to explain consumers’ intentions to contribute product reviews to OFSs.

3.1 Research hypotheses

Number of existing reviews Benefits

„

„

Intrinsic Enjoyment in helping other consumers Enjoyment in influencing the

H1a H1 H2a H2 H3a

Internalized Extrinsic „ Self-enhancement

H3

Intention of Information Contribution to OFSs

H4a „

„ „

Extrinsic Economic reward

Costs Cognitive cost Executional cost

H4

H5 H6

Figure 1. Research model

3.1.1. Enjoyment in helping other consumers (EHLP). This benefit is closely related to the concept of “altruism” in philosophical literature. People who are motivated by altruism perceive intrinsic satisfaction from attempting to optimize other people’s utilities without expectation of return from others [39]. This situation is especially likely to happen when people perceive that their behavior would be beneficial to others [14]. Providing an online product review can help other consumers to develop comprehensive knowledge of the product which is unlikely to obtain from advertisement [46]. Therefore, consumers may choose to share their own positive or negative consumption experiences in OFSs as a means to help other consumers [25]. These individuals gain pleasure by aiding others to make a satisfying purchase [42]. Crowding effect is often associated with altruistic behavior [2, 11]. In the context of OFSs, a small number of existing reviews implies the possibly higher utility of a new review in helping other consumers to make better purchase decisions. Hence, people can get more supportive feeling and pleasure under this condition and their intentions to write the review increase. Thus, we hypothesize: H1: Enjoyment in helping other consumers is positively related to intention of information contribution to OFSs.

3

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

H1a: The positive relationship between enjoyment in helping other consumers and intention of information contribution to OFSs will be stronger under small number of existing reviews. 3.1.2. Enjoyment in influencing the company (ECOM). EWOM literature suggests that the enjoyment in influencing the company may be an intrinsic benefit that consumers can gain from providing product reviews in OFSs. The enjoyment arises from two sources: helping the company and vengeance [25,42]. Enjoyment in helping the company is a benefit which is also based on altruism. However, the object involved here is the company that provides a good product [42]. Due to satisfied consumption experiences, the consumer may have a desire to give good company something in return [25]. Enjoyment in vengeance is associated with negative consumption experiences. When consumers perceive that the company provides bad products, they may develop a desire to retaliate against the company [42]. To the extent that EWOM plays an important role in shaping consumers’ buying decision [6], people are able to exert power to influence companies’ businesses by giving recommendations/warnings in OFSs [25, 43]. Therefore, consumers are able to gain pleasure in OFSs by providing reviews to influence the company [42]. Reasoning from MCT, we expect that when the number of existing reviews is large, consumers who want to influence a company may feel they are supported by others who already write reviews for the same product. As the collective power can be enlarged by each individual’s contribution, the probability of influencing the company’s business will correspondingly increase. Therefore, the influence of a consumer’s intention will be heightened when the existing number of reviews is large. Thus, we postulate, H2: Enjoyment in influencing the company is positively related to intention of information contribution to OFSs. H2a: The positive relationship between enjoyment in influencing the company and intention of information contribution to OFSs will be stronger under large number of existing reviews. 3.1.3. Self-enhancement (SLFE). Self-enhancement is defined as the tendency to seek experiences and feedbacks from others that improve or bolster the status [47]. The self-enhancement effects are much stronger in OFSs than offline WOM because the nature of OFSs allows consumers to provide their reviews to

a larger amount of people and keep reviews for a long period [29]. In the EWOM domain, self-enhancement has been identified as a possible benefit for consumers to provide reviews [14, 42, 47]. It has been found that consumer participation is employed as a means to gain reputation to other consumers [14, 21]. To the extent that OFSs provide much larger room for consumers to gain a kind of connoisseurship from other people compared to offline WOM, self-enhancement expectancy is expected to motivate consumers to provide post product reviews to OFSs. However, the effect of self-enhancement is conditional on the number of existing reviews. Most OFSs award reviewers’ reputation based on quantity and quality of reviews written. The first few reviews of each product will be given higher weight in computation of status as they are more valuable to other consumers. In the presence of a large number of existing reviews, the consumers may find it harder to obtain recognition as their reviews will be given low weight. Thus, we hypothesize that: H3: Self-enhancement is positively related to the intention of consumer information contribution in OFSs. H3a: The positive relationship between selfenhancement and the intention of information contribution to OFSs will be weaker under conditions of large number of existing reviews. 3.1.4. Economic reward (RWAD). Previous research has shown that monetary reward is a positive extrinsic motivator to increase people’s level of participation and performance [37, 45]. Many famous OFSs provide economic reward to its reviewers under conditions tied to their review performance, usually gauged by the appraisal from other consumers. However, the positive relationship between economic reward and the intention of contribution is likely to be moderated by number of existing reviews. Consumers often want to obtain useful information about the latest products. Therefore, the earliest reviews of a product are likely to be valuable and have higher chance to be rated as “very helpful”. In order to earn more economic reward, consumers should choose a product which is unreviewed previously but somewhat popular 1 . Practically, to be among the first to write about a product is always the most important criterion to earn more reward in many OFSs. Thus, we hypothesize that: 1

For more information on criteria of gaining reputation in OFSs, please visit http://www.epinions.com/help/faq/show_~faq_recognition

4

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

H4: Economic reward is positively related to the intention of information contribution in OFSs. H4a: The positive relationship between economic reward and the intention of information contribution to OFSs will be weaker under conditions of large number of existing reviews. 3.1.5. Cognitive cost (COGC). Psychologists state that individuals need to process extensive information cognitively before physical response to environmental stimulus [15,36]. For example, in order to integrate information, people need to retrieve further information from memory. The act of review contribution to OFSs requires consumers to share their past consumption experience relating to the product. They need to recall detailed past consumption experience and mentally organize each piece of knowledge, perception and evaluation into a detailed product review. This complex cognitive process may yield negative psychological states such as annoyance and unpleasantness, which in turn would lead to intrinsic cost of contributing product reviews. The intrinsic cost would decrease the intention to provide reviews. Therefore, we hypothesize, H5: Cognitive cost is negatively related to the intention of information contribution to OFSs. 3.1.6. Executional cost (EXEC). By contributing detailed product reviews in OFSs, consumers need to materialize or codify their tacit opinions into the online system. The time demand for inputting reviews in OFSs is substantial when providing detailed review. This process is considered as contribution cost because the large time and effort foregone can be used to do other alternative behavior and to obtain the corresponding rewards or benefits accruing from that behavior [4,28]. Previous research shows that when the knowledge contribution requires significant time, knowledge sharing tends to be inhibited [35]. Therefore, we hypothesize, H6: Executional cost is negatively related to the intention of information contribution to OFSs. 3.1.7. Control variable. Previous literature shows that opinion leadership (OPLD) may influence people’s intention to participate in WOM [14]. Opinion leader is the person from whom a consumer often seeks consumer-related advice [16]. We consider it as a control variable to ensure that it does not interfere with the study focus.

4. Research methodology 4.1. Research design Experimental design was adopted for this study. In this study, two constructs were manipulated, namely the economic reward (present or absent) and the number of existing reviews for a product (small or large). Therefore, a 2u2 between-subjects factorial design was used in our experiment and totally four scenarios were generated. Economic reward was chosen to be manipulated because this reflects the reality of current OFS operation. In our study, economic reward was presented as shopping vouchers that are awarded to reviewers under conditions that were designed in accordance with ePinions.com’s income share program. Two numbers, 0 and 30 were used to reflect the small and large number of existing reviews respectively.

4.2. Instrument development The instrument development procedure was based on Churchill’s [7] framework. Measures that had been validated in previous researches were adopted to enhance the validity of measures and adapted to the current topic [28]. 37 items were generated for eight constructs as the outcome of the first two steps of instrument development. All measurements used 7point numeric scale, with 1 being strongly disagree, 4 being neutrally agree and 7 being strongly agree. Sorting procedures were used to ensure conceptual validation where items at operational level reflected the constructs at the conceptual level [31]. Two parts of sorting: unstructured and structured were used to ensure the pertinence of each question to its own construct. Cohen’s Kappa and item placement hit ratio were selected to assess the result of each sorting stage. Modifications were made according to each sorting results and judges’ comments. 27 undergraduate students were voluntarily participated in the pilot test. Cronbach Alphas and Factor analysis with varimax rotation were computed to test reliability of items and instrument validity. Final questions for this study are listed in Appendix A.

4.3. The experimental task Undergraduate students from a large university in Singapore were selected to be participants in this experiment as they were believed to be more familiar with online communities. The subjects were chosen from all faculties in order to maximize the generalizability of our results and avoid selection bias.

5

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

Study booklets that contained experiment manipulations and questions were distributed to the students after lectures or tutorial classes. Their participation was voluntary but they were given a lucky draw incentive. The completed booklets were either collected by experiment investigators in class or returned directly by the subjects to a box outside investigator’s office. The experiment booklet began with a cover letter stating the purpose of this study. The experiment was based on a specific OFS, the Amazon feedback system. A short introduction to this system was given and followed by a description and a sample of detailed product review. The subjects were randomly assigned to one of the four scenarios (each representing different level of reward and existing reviews) and required to answer some questions for manipulation check. At the end of experiment, some demographic information was collected while the subjects were assured of the confidentiality of their information. They were asked to return the completed experiment before the date of lucky draw. Out of 400 questionnaires sent out, 188 responses were received resulting in an overall 47% response rate. 176 entries were recorded as 12 returned questionnaires were incomplete and removed.

5. Data analysis The summary of demographic information from experiment is shown in Table 1. All statistical tests were assessed at 5% level of significance.

5.1. Manipulation and control checks Prior to statistical testing, manipulation and control checks were performed. Several questions were asked in the questionnaire for manipulation check. Results showed that all scenarios were manipulated successfully. Control checks on demographic data and opinion leadership construct in each treatment group were performed. Multivariate tests on age (F=0.397, p=0.530), starting years of using the Internet (F=3.393,p=0.067), and opinion leadership construct (F=1.537,p=0.211) and Mann-Whitney U tests on the subjects’ gender and experience of using OFS did not differ across the four scenarios.

5.2. Measurement model The reliability was tested using Cronbach Alpha. Nunally [34] stated that Cronbach Alpha greater than 0.7 indicated adequate reliability. Convergent validity was assessed through three methods: item reliability,

composite reliability of constructs, and average variance extracted (AVE) [17]. A score of 0.5 indicated acceptable level of variance extracted and the two reliability scores should be at least 0.707. Results of these tests are presented in Table 2. For our study, all the data exceeded the acceptable. Therefore, good reliability and convergent validity were achieved. Table 1. Demographic Information Scenario

Frequency Combined (176) Reward*Large existing reviews (44) Reward*Small existing reviews (47) No reward*Large existing reviews (43) No reward*Small existing reviews (42)

Gender

Age

M F

Mean (std dev)

Year starting using the Internet Mean (std dev)

Prior experience using OFSs Yes No

92 84 52.3% 47.7% 59.1% 40.9%

--

--

21.61 (1.67) 21.55 (1.55)

1997.92 (2.24) 1997.34 (3.11)

26 150 14.8% 85.2% 11.4% 88.6%

66% 34%

21.47 (1.28)

1998.04 (1.61)

14.9% 85.1%

37.2% 62.8%

21.56 (1.99)

1998.42 (1.735)

16.3% 83.7%

61.9% 38.1%

21.88 (1.82)

1997.88 (2.14)

16.7% 83.3%

Table 2. Assessment of Reliability and Convergent Validity Dimension

Cronbach’s Alpha

Composite Reliability

ITEN EHLP ECOM SLFE COGC EXEC OPLD

0.959 0.955 0.897 0.947 0.916 0.944 0.917

0.973 0.967 0.928 0.962 0.941 0.960 0.941

Average Variance Extracted (AVE) 0.924 0.881 0.762 0.862 0.799 0.856 0.799

Table 3. Assessment of Discriminant Validity ITEN

EHLP

ECOM

ITEN

0.924

EHLP

0.647

0.881

ECOM

0.418

0.471

0.762

SLFE

0.140

0.178

-0.145

COGC

-0.346

EXEC

-0.319

OPLD

0.318

0.391 0.232 0.415

0.480 0.186 0.332

SLFE

COGC

EXEC

OPL D

0.862 0.082 -0.006 0.083

0.799 0.324 -0.313

0.856 0.324

0.799

Discriminant validity was assessed by two methods: factor analysis and item correlation [17]. The

6

Proceedings of the 40th Hawaii International Conference on System Sciences - 2007

commonly accepted threshold for item loading was 0.5, while loadings above 0.71 were considered excellent. The results yielded eight components with eigenvalues greater than 1. All items loaded higher on their targeted constructs and exceeded the accepted threshold. The second method to test discriminant validity suggested that the AVE of both constructs should be greater than the variance shared by the two (i.e. the squared correlation) [17]. Table 3 reports the result of this method. All items fulfilled the requirement of discriminant validity.

5.3. MRA and its Assumptions Multiple regression analysis (MRA) is an advanced statistic technique for simultaneous assessment of the relationship between a single dependent variable and several independent variables [23]. Moreover, the moderating effect can be tested by multiplicative terms using an extension of MRA (MMR). As suggested by Hair et al. [23], the overall significance of our model is accessed by F ratio and R2 (coefficient of determination). Before testing hypotheses, some assumptions were checked [23]. The normality of variable can be tested using skewness and kurtosis. The value between -2.5 and 2.5 represents a normal distribution. Outliers are the cases which are more than plus or minus 3 standard deviations from the mean value. Multicollinearity can be tested using tolerance value and variance inflation factor (VIF). The acceptant range is tolerance > 0.2 and VIF