• 沒有找到結果。

Examining the effect of information disclosure on website stickiness: The cognitive information perspective

N/A
N/A
Protected

Academic year: 2021

Share "Examining the effect of information disclosure on website stickiness: The cognitive information perspective"

Copied!
33
0
0

加載中.... (立即查看全文)

全文

(1)

Examining the effect of information disclosure on

website stickiness: The cognitive information perspective

從資訊認知觀點探討資訊揭露對網站黏著度的影響

Tser-Yieth Chen

Graduate Institute of International Business, National Taipei University

Fang-Ming Hsu1

Department of Information Management, National Dong Hwa University

Shih-Yin Wu

Graduate Institute of International Business, National Taipei University

Abstract:

This study employs the concept of cognitive information processing (CIP) to examine the causal relationships among information disclosure, cognitive information effect, and website stickiness. Considering their interests, consumers may become “sticky” to a specific website after recognition of some exposed information. We thus look to examine the cognitive information effect, i.e. information scent and message framing, in consumers’ processing of their perceived information. The findings from an empirical investigation demonstrate that sponsorship disclosure significantly affects the information scent, which mainly positively impact the website stickiness. Moreover, sponsorship disclosure positively influences message framing, which then positively affects website stickiness. Our findings indicate that sponsorship disclosure plays a main role in forming website stickiness. This research marks a new ground in the sponsorship disclosure and endorsement disclosure literature and thus provides a useful profile in the field of on-line marketing. Marketing managers can hence use appropriate information disclosure to attract and stick consumers to a site by way of information scent and message framing.

Keywords:

Information disclosure, message framing, information scent, stickiness, e-WOM.

1

Corresponding author: Fang-Ming Hsu, Department of Information Management, National Dong Hwa University, No.1, Section 2, Da Hsueh Road, Shoufeng, Hualien County, 97401, Taiwan, ROC. E-mail: fmhsu@gms.ndhu.edu.tw.

(2)

摘要:本研究運用資訊認知過程概念,檢視資訊揭露、資訊認知效果和網站 黏著度之間的因果關係。考慮到消費者的興趣,消費者在認知到若干揭露的 資訊之後,可能會對特定網站產生黏著。因此,本文藉由資訊嗅覺和訊息框 架等資訊認知效果,檢視在消費者處理他們資訊認知的反應結果。實證結果 顯示,贊助者揭露對資訊嗅覺有顯著影響,資訊嗅覺對網站黏著度也有正向 顯著影響。此外,贊助者揭露對訊息框架有正向影響,進而對網站黏著度也 有正向顯著影響。此研究結果顯示,贊助者揭露在網站黏著度的形成過程中 具有主要的作用。本研究是贊助者揭露和名人代言文獻的一個新領域研究, 有助於提供網路行銷領域相關的應用。因此,行銷經理可以使用適當的資訊 揭露,透過資訊嗅覺和訊息框架的方式,吸引和黏著消費者到特定網站中。 關鍵字:資訊揭露、訊息框架、資訊嗅覺、黏著度、e 化口碑

1. Introduction

The stickiness of customers to continuous visiting is a critical focus for website managers when conducting marketing on the Internet (Chou and Chen, 2016). Stickiness is also the key to attracting consumers and important for many industries (Hwang and Jeong, 2016), and it could be increased through interactive and vivid content that encourages consumers to revisit and spend a longer period of time on the website (Lin, 2007). In this information era, customers are always searching and reviewing related information in order to make suitable decisions and exhibit rational behavior (Cheng, Chien, and Woodburne, 2017; Ho and Bodoff, 2014). Moreover, stickiness is the process of converting information into action (Kurstedt, 1987), and converting decisions into an action is the necessary result from using information.

Information is encoded by consumers to give meaning and can be compared with their own interests. After the information is collected and decisions are made, consumers start to take action. Interactions among groups and information systems produce specific symbolic meanings that satisfy self-valued performance (Hsu et al., 2013). Thus, information is the foundation and starting point of consumption behavior, with appropriate information being a precondition for effective decision making and action. This study thus uses the concept of information processing theory, which reduces uncertainty in consumers’

(3)

decision-making behavior, to explore the cognitive information effect on their website stickiness.

Users have a higher sticky intention with a message when they encounter an interesting message (Gibson et al., 2016). Information disclosure is a managerial tool of Internet marketing, and endorsement and sponsorship disclosure are clear signals from a website trying to build social ties based on the signal mechanism. This can lead to consumers’ trust in the sponsorship and follow-up behaviors (Mudambi and Schuff, 2010). These signals are traits that have evolved specifically, because they can change the behavior of consumers in ways that benefit the signal provider. Marketers use the mechanism of signaling to disclose details of schedules and to give the latest authorization, which have been shown to be more effective and efficient in innovative website development (Moody and Galletta, 2015).

The relationships among information scent, messaging framing, and stickiness are also a significant issue for online shopping platforms, especially as the popularity of the Internet has grown dramatically (Moody and Galletta, 2015). Numerous research studies have discussed sponsorship, but there are limited studies exploring the relationship between information disclosure and cognitive information cues of websites during the process of cognitive information processing (CIP) (Boerman et al., 2017). Surveys regarding the effects of message framing and information scent on website stickiness are also rare and form the research gap this study is targeting to fill. Consequently, we shall examine the causal relationships among information disclosure, the cognitive information effect of cues, and website stickiness.

2. Literature review and development of hypotheses

The concept of information processing is based on the idea that humans process the information they receive, rather than merely responding to stimuli. Pfeffer (1981) provided a particularly clear summary of this. First, the individual’s environment may provide cues as to which dimensions might be used to characterize the work environment. Second, the environment may provide information concerning how the individual should weigh the various dimensions - whether autonomy is more important than variety of skill, whether pay is more or

(4)

less important than usefulness or worth, etc. Third, the context provides cues concerning how others have come to evaluate the work environment on each of the selected dimensions. Fourth, it is possible that the context provides a direct evaluation of the work setting along with positive or negative dimensions, leaving it to the individual to construct a rationale to make sense of the generally shared affective reactions. In this study, a website is the environment that provides consumers information for their decision making behaviour. Websites need to disclose relevant information that consumers can use to diagnose the necessary unmet demand of suggestive details.

Based on CIP, endorsement and sponsorship disclosure may be used by customers to reach the website of their interests. Endorsement disclosure is defined as any individual who uses public recognition on behalf of a product by appearing with it in an advertisement (McCracken, 1989). The visual information is transformed as a series of processing stages in the individual memory systems. To equate endorsers’ good appearance with the image of the products, advertisers use endorsers to shape the brand image of products, especially when there lacks any product image definition (Walker et al., 1992).

Sponsorship disclosure is a notification that indicates that the content was sponsored, so that consumers recognize the persuasive intent of the content from the CIP concept (Boerman et al., 2012; 2014b; 2017). Sponsorship disclosure is also a cue that consumers use to determine whether or not a product is worthwhile, according to the information diagnosis mechanism (Ohanian, 1991). This content attracts consumer attention and can then serve as helpful information for evaluating the usefulness of the sources.

Information scent refers to a seeker’s evaluation of proximal and informational cues such as website links and other navigational aids to assess the relative likelihood of finding valuable and potentially desired information on other pages of the website (Rajala and Hantula, 2000). The information processing progression is assumed to be cognitive and learned. Information scent refers to a consumer’s evaluation of proximal, imperfect, and informational cues to assess the relative likelihood of finding valuable and potentially desired information located on other pages within the website (Rajala and Hantula, 2000). The consumer assesses the “scent” provided by the website’s text and determines whether the information leads to desired content. Thus, information scent reduces

(5)

the cost of information retrieval by indicating whether browsing a given area of the website, continuing to a new area, or moving to a different website is desirable (Moody and Galletta, 2015).

Message framing is the way in which CIP is presented (Chong and Druckman, 2007), of which there are two types. In the first type, the message conveys the benefits of using the product while emphasizing the opportunity cost of not using the product (Overton, 2018). In the second type of message framing, positive messages are connected with the positive attributes of the product or claim and negative messages with consumers’ low trust and their nature of advertisement (Meyers-Levy and Maheswaran, 2004). Sponsorship disclosure and endorsement can influence the reaction of the audience given the information diagnosis mechanism within the CIP concept (Boyd and Ellison, 2008). Since information diagnosis occurs between providers and consumers, providers need to disclose credible information that consumers can use to judge the potential benefits and opportunities of suggested projects (Boerman et al., 2017).

Website stickiness is the ability of a website to gain continuous consumers by attracting passersby to stop and then continue to use the website and is deemed as the loyalty to the website (Lin, 2007). Amusing and interactive content positively affect a consumer’s attitude towards social networking sites, especially when information scent and message framing suggest usefulness, thus forming a stickiness intention to remain involved (Wu, 2011).

Electronic word-of-mouth (e-WOM) is defined as any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet (Henning-Thurau et al., 2004). Before the advent of information technology, e-WOM meant consumers spreading their experience among friends and relatives via online media (Viglia, Minazzi, and Buhalis, 2016). E-WOM has an impact on customer attitudes and consequently on booking/purchase intentions (Parket et al., 2007). Therefore, e-WOM is the critical factor that website managers want to keep for further marketing.

It is important to recognize cognitive information that influences stickiness so that marketing managers can maximize the volume of information scent and message framing (Mudambi and Schuff, 2010). Individuals use the endorsement and sponsorship disclosures to evaluate the information by attribution or labeling

(6)

H5 H1a H2a H4 H3a H1b H2b H3b

within CIP. The learning process must be directed toward the achievement of common goals (Chen, Huang, and Wey, 2017). Attribution and labeling are closely related to social construction and symbolic interaction analysis in the cognitive learning process. Using attribution and labeling, an information seeker locates cognitive information more efficiently, rather than committing substantial cognitive resources to determine whether to continue searching in a given location that may or may not contain the desired information (Moody and Galletta, 2015). They then involve the application mechanism of the CIP effect of website cues. Our research is based on the study of Rothman (2005) and Gibson et al. (2016) and employs the CIP concept further by providing practical applications for better marketing. The cognitive information effects, such as information scent and message frames, on the website are included in cognitive communication. Figure 1 illustrates the research framework of this study.

Figure 1 Research framework e-WOM Information Scent Website Stickiness Endorsement Disclosure Sponsorship Disclosure Message Framing Information Disclosure Cognitive Information Effect E-loyalty

(7)

2.1 Effect of endorsement disclosure on information scent and message framing

Via social learning from the CIP concept (Mudambi and Schuff, 2010), endorsements create a distinctive personality for a brand, powerfully creating a positive image that consumers could imitate. When people have strong social ties, the information to him allow a large number of people to have access to this message (Keng, Chen, and Huang, 2018). Through endorsements, marketers provide social ties via official media certified by third parties, and consumers use these social ties to enhance interactions. Endorsements reduce the cost of information retrieval and determine whether a consumer will search for information with similar scents to the endorsement (Moody and Galletla, 2015). Celebrity endorsement disclosure offers positive attributes of the product or claims a connection through positive messages. Thus, the consumer will transfer his good impression of the celebrity to the product being endorsed if the consumer is the fan of the celebrity and has a good impression of the celebrity (Meyers-Levy and Maheswaran, 2004). Through the halo effect, the endorsement can be deemed as an information scent to guide consumers’ attention to the advertisement, arouse their interest, and pique their desire. Thus, we propose the following hypothesis.

H1a: Endorsement disclosure positively influences information scent.

Celebrity endorsement disclosure can provide a brand image cue to attract consumers’ attention in order to form resonance and promote a subsequent purchase. Given that the image principle is used, the brand image cue from celebrity endorsement disclosure can deduce or maintain the product evaluation (Kamins et al., 1989). Thus, when brand image is high, its overall evaluation is high. This will form positive message framing and also make the perception of quality high as well (Dodds et al., 1991). Endorsement disclosure may propose a connection between consumers’ low trust and low preference, since consumers will deem celebrity endorsement as another type of advertisement and form negative message framing based on the information diagnosis mechanism (Chong and Druckman, 2007; Ohanian, 1991). Thus, we propose the following hypothesis.

(8)

2.2 Effect of sponsorship disclosure on information scent and message framing

Sponsorship disclosure provides an information cue to help consumers find a linkage of positive branding, i.e., a type of information scent (Kerne and Smith, 2004; Stipp and Schiavone, 1996). Sponsorship cues influence consumers who depend on the attention to both the information source and the information itself (Boerman et al., 2017). The high reputation of a sponsor draws consumers’ attention. Consumers recognize sponsorship as a link to products, and this phenomenon renders the effect of information scent (Boerman et al., 2012). Therefore, sponsorship disclosure is an easy way to raise brand reputation and has a real effect on the endorser. Thus, we propose the following hypothesis.

H2a: Sponsorship disclosure positively influences information scent.

Even while consumers may prejudge sponsorship as the nature of an advertisement to form negative message framing, sponsorship disclosure can propose a connection between consumers’ trust and preference (Meyers-Levy and Maheswaran, 2004). Reading an advertisement without disclosure of sponsorship is equivalent to online reviews without origin. Not knowing the source of the information, consumers lose the option to willingly accept sponsorship information, and sponsors lose their ability to invite consumers through the content (Nelson et al., 2009). In other words, companies utilize sponsorship activities to increase the visibility of their products and services and thereby their sales and profitability (Sandler and Shani, 1993). This will affect consumers’ cognition when they have a positive evaluation of message framing. Since it provides a quality assurance from the sponsorship of a famous company, sponsorship disclosure can promote products’ positive attributes or claim a connection via positive messages (Overton, 2018). Thus, we propose the following hypothesis.

H2b: Sponsorship disclosure positively influences message framing.

2.3 Effect of information scent and message framing on website stickiness

The development of information scent requires both collecting timely information about what consumers find important and constantly testing

(9)

assumptions of their expectations (Spool et al., 2004). Along with the self-concept theory, consumers must find cues that are consistent with their self-concept, e.g. self-image, to reach their self-persuasion (Johar and Sirgy, 1991). The more advertising messages are in line with the self-concept of consumers, the stronger the persuasion becomes. Website stickiness with a given unit of information could be measured as the incremental expenditure required to transfer that unit of information to a specified locus in a form usable by a given information seeker (Wu, 2011). When this expenditure is low, stickiness is low; when it is high, stickiness is high. Under self-persuasion, the consumer enters a continuous state of stickiness (Kwon and Sung, 2012). Consumers expect to buy products when the perception of informed products is consistent with their expectations, and when product perception is inconsistent with their expectations, it leads to a low level of consumers’ interest (Churchill and Surprenant, 1982). Therefore, we propose the following hypothesis.

H3a: Information scent positively influences stickiness.

Message framing is the way in which CIP is presented. Along with the cognitive response theory, when consumers receive the trait of a message, they evaluate the message and react appropriately (Greenwald, 1968). Consumers always prefer attractive things, and hence the attractiveness of a product affects its persuasion. The impact of attractiveness on consumers’ attitudes comes through consumers’ meaning and recognition to identify the attractiveness of the source (Overton, 2018). Under message framing, the usefulness and interaction of information have a positive impact on consumers’ attitude towards social networking sites, thus forming a stickiness intention to remain as a participant (Wu, 2011). Thus, we propose the following hypothesis.

H3b: Message framing positively influences stickiness.

2.4 Effect of information scent on message framing

Recreating the linkage of information scent increases message framing (Nisbet, 2009). Once consumers detect useful information and choose further actions, they then decipher whether the information is positive or negative and judge its benefits and opportunities. As soon as consumers recognize their own needs and motivate themselves to feel the necessity to take actions, they consider this message framing as a resultant message and render a positive attitude

(10)

(Boerman et al., 2014a). When exposed to an information scent, consumers’ interpretation naturally connects their concept between their heart and the frame of things discourse, and they then become the message constructor (Moody and Galletta, 2015). Message framing is effective as a consequence of information scent. Thus, we propose the following hypothesis.

H4: Information scent positively influences message framing.

2.5 Effect of website stickiness on electronic word-of-mouth (e-WOM)

Stickiness is the ability to keep consumers engaged in community activities repeatedly for long periods of time (Zott et al., 2000). If a community has a high degree of stickiness, then consumers in that community are loyal to it (Holland and Baker, 2001). Loyal consumers share their satisfaction in the community with their relatives and friends and recommend the community to others on the Internet (Weisfeld-Spolter, 2014). Thus, we propose the following hypothesis.

H5: Website stickiness positively influences e-WOM.

3. Research method

This study employed a questionnaire with a 6-point Likert scale, including endorsement, sponsorship disclosure, information scent, message framing, and e-WOM. Endorsement has three categories: Attractiveness, self-expression, and relevance to life (McIntyre and Pigram, 1992). Sponsorship disclosure has three categories: Persuasion knowledge, timing, and duration (Boerman et al., 2014a). Information scent has two categories: inferring the consumer’s need and stimulating consumer action (Gibson et al., 2016). Message framing has four categories: The positive and negative of the message itself, the positive and negative of message attribution, the deterministic message of benefit, and the deterministic message of opportunity (Rothman, 2005). Stickiness has three categories: Length of time, frequency, and depth of visit (Lin, 2007). Finally, e-WOM has three categories: content, consistency, and volume (Viglia et al., 2016). Survey respondents were asked their opinions about a particular website of a designer and branded bag store with physical/online stores.

With the goal of discovering stickiness and consumers’ e-WOM behaviour, this study examines data from consumers with online shopping experiences in

(11)

Taiwan, using quota sampling via a website survey. We employ quota sampling to assure that the classification of the entire population is included in the selected samples. Quota sampling confirms that the population characteristics of a sample match those of the entire population (Cooper and Emory, 1995). We also take age, gender, and region to segment the samples since these three are popular demographics used as segmentation variables.

For people above 25 years old, the population of broadband Internet in Taiwan is estimated to be 13.15 million people, accounting for 57% of the total population (Taiwan Network Information Center, 2017). We determine what ratio of the population has engaged in online use according to gender, age, and region from a survey of Taiwan’s online users by the Taiwan Network Information Center. We then determine the sample structure and how many samples in each cell we need. Among online consumers, males account for 49.8%, and females account for 50.2%. We divide our samples into three different age groups, 18-49 years (47.2%), 50-64 years (32.8%), and 65 and over (20.0%). We exclude those under 18 years since their income does not allow for purchasing expansive 3C products. We further divide our samples into four different regions of Taiwan: northern, central, southern, and eastern (and outer islands). Online consumers in the northern region account for 44.9%, consumers in the central region are 24.6%, consumers in the southern region are 27.4%, and consumers in the eastern region are 3.1%.

We next utilize stratified sampling, dividing the population into several online user groups based on the 2013 population statistics of Taiwan. Thus, the designed sample structure between males and females is 159 (49.7%) and 161 (50.3%), respectively. As to age, 18-49 years old are 151 (47.2%), 50-64 years old are 105 (32.8%), and above 65 years old are 64 (20.0%) in total. The population structure of online users among different regions in Taiwan is: northern Taiwan 144 (44.9%), center Taiwan 79 (24.6%), southern Taiwan 87 (27.4%), and eastern Taiwan and outer islands 10 (3.1%). Table 1 shows the sample structure.

To discover and eliminate potential problems in this study, we employ a pre-test to make sure of the validness of measurement adapted herein before carrying out a formal questionnaire. We collect thirty questionnaires to run the pre-test. The values of Cronbach’s α of each variable, including endorsement

(12)

behaviour, sponsorship disclosure, information scent, message framing, stickiness, and e-WOM, are 0.697, 0.696, 0.838, 0.844, 0.896, and 0.803, respectively. All the values are over 0.6, thus passing our criteria of reliability analysis, and so we can say this pre-test exhibits internal consistency.

We then collect 320 samples in the formal test, categorized by gender, age, area, education level, and income level in Table 2. First, the actual samples between male and female are 160 (50%) and 160 (50%), respectively. Second, our samples within the three different age groups are 18-49 years (156, 48.75%), 50-64 years (97, 30.31%), and 65 and over (67, 20.94%). In addition, empirical samples by region are as follows: northern (142, 44.38%), central (79, 24.68%), southern (91, 28.44%), and eastern (8, 2.50%). We conduct ANOVA analysis to examine whether the sample characteristics will affect the result (e-WOM) and to present the sample as being representative or not. We adopt gender, age, region, and education as the characteristics that may affect the results. The p-values of 0.579, 0.844, 0.226, and 0.565, respectively (p > 0.05), for gender, area, region, and education level are not significant at affecting e-WOM.

Table 1 Sample structure Items 18-49 years-old 50-64

years-old Over 65 years-old Total (%) Male Female Male Female Male Female

North Taiwan 34 34 23 24 14 15 144 (44.9%) Central Taiwan 19 18 13 13 8 8 79 (24.6%) South Taiwan 20 21 14 15 9 8 87 (27.4%) East Taiwan & Islands 3 2 1 2 1 1 10 (3.1%) Sub-total 76 75 51 54 32 32 320 Total 151 (47.2%) 105 (32.8) 64 (20.0) 320 (100.0%)

(13)

Table 2

Demographics of sample

4. Empirical result and testing of the hypotheses

We also use a pre-test to verify the validity of our measurements. We collect 30 questionnaires to run the pre-test, which checks the survey wording and meaning. The Cronbach’s alpha values for the constructs of endorsement behavior, sponsorship disclosure, information scent, message framing, stickiness,

Characteristics Category Count Percentage (%)

Gender Male Female 160 160 50.00 50.00

Age 18 ~ 49 years old

50 ~ 64 years old Above 65 years old

156 97 67 48.75 30.31 20.94 Education

Below Junior high school Senior high school College Master Ph. D 50 57 185 26 2 15.63 17.81 57.81 8.13 0.62 Occupation Student Government Manufacturing Commerce Service industry Information technology Transportation and warehousing Self-employed profession Retirement Other 32 10 17 61 80 6 5 34 51 24 10.00 3.13 5.31 19.06 25.00 1.88 1.56 10.63 15.93 7.50

Region Northern Taiwan

Central Taiwan Southern Taiwan

Eastern Taiwan and islands district 142 79 91 8 44.38 24.68 28.44 2.50

(14)

and e-WOM in the pre-test are 0.697, 0.697, 0.838, 0.844, 0.896, and 0.803, respectively. As these values are greater than .6, they are within the reasonable standard of internal consistency and reliability (Nunnally, 1978).

We jointly assess the reliability for all items of a construct by computing the composite reliability (CR). According to Fornell and Larcker (1981), if the composite reliability is larger than 0.6, then this indicates an acceptable fit of the data. Consequently, the higher the composite reliability (CR) of latent constructs is, the easier we might examine the latent constructs. In our samples, the calculated values of CR for endorsement behavior, sponsorship disclosure, information scent, message framing, stickiness, and e-WOM are respectively 0.917, 0.816, 0.906, 0.939, 0.884, and 0.884. All of the values are over 0.6, surpassing our criteria of reliability analysis, and so we say that this test illustrates internal consistency. The results of composite reliability of each variable are in Table 3. We also use Cronbach’s alpha coefficient to examine the internal consistency reliability of each construct. Cronbach’s alpha coefficients of all items of the six constructs from customer samples are greater than 0.7 and reveal high internal consistency (Bagozzi and Yi, 1988; Nunnally, 1978).

We additionally calculate the average variance extracted (AVE) to confirm the discriminant validity in our study. If average variance extracted is larger than 0.5, then discriminant validity between those variables is achieved (Fornell and Larcker, 1981). We calculate the average AVE of endorsement behavior, sponsorship disclosure, information scent, message framing, stickiness and e-WOM as 0.783, 0.750, 0.824, 0.840, 0.736, and 0.800, respectively. We find that all AVE are larger than 0.5 and reveal that the variables can correctly measure what they mean and hold discriminant validity.

In our model we adopt multiple fitness indices to examine the validity of the model. Table 4 lists the fit indices of the proposed measurement model. First, we calculate χ2 /df. is 3.499, which fits the criteria 2<X<5 (Anderson and Weitz, 1992). Second, GFI represents the explained variance and covariance by the model (Jöreskog and Sörbom, 1993). The higher GFI is, the greater the variance that can be explained by the model. We see 0.869 for GFI, but this does not pass the criteria of >0.9. AGFI is 0.816. Note that we can adopt CFI (0.933) and NFI (0.910) to judge the data’s validity, because these indices are above 0.9 (Anderson and Weitz, 1992). We further use RMR and RMESA to provide

(15)

Table 3

Results of reliability and validity analysis

Construct Item Loading Cronbach’s α CR AVE

Endorsement Attractiveness of endorser Self-expression of endorser Relevance to life .885 .942 .843 .917 .915 .783 Sponsorship Disclosure

Persuasion knowledge of sponsor Timing of sponsorship Duration of sponsorship .828 .714 .793 .816 .900 .750 Message Framing Positiveness of message itself Positiveness of message attribution Determination of message: benefit Determination of message: opportunity

.782 .845 .905 .849 .906 .949 .824 Information Scent Relevance to user’s need Trigger of user’s action

.929 .950

.939 .913 .840

Stickiness Duration of stay on website

Frequency of visiting website Depth of visit in website

.838 .934 .799 .884 .893 .736 e-WOM Richness of content Consistency of recommendation Volume of online review

.880 .897 .788

.885 .923 .800

information about the fit of the model with unknown but optimally chosen parameter values for the population covariance matrix, if it is available (Bagozzi and Yi, 1988). In our study, RMR is 0.03, which is less than 0.05. RMSEA is 0.091. The results of RMR and RMSEA indicate a good fit.

Bagozzi and Yi (1988) referred to various criteria to judge the goodness-of-fit of the structural model. Only one criterion (say, GFI and/or AGFI must exceed 0.9) may be too conservative to judge the model, because the number of measurement items obviously influences the magnitude of GFI and AGFI (Bagozzi and Yi, 1988). The more measurement items there are in the

(16)

model (such as 54 questions in this study, excluding demographic variables), the less easy it is to reach the 0.9 overall suitability (GFI) evaluation criteria. Bollen (1989) argued that aside from GFI and AGFI, it is recommended to adopt CFI and NFI criteria in order to make the variation more stable; if these two criteria are greater than 0.9, then the study’s model fits well. When the number of samples is larger, CFI is superior to GFI (Hu and Bentler, 1999). In addition, if a single criterion (such as this study’s RMSEA) fails to pass the standard, then one should combine and refer to multiple criteria from different dimensions to test the model and make a reasonable judgment (Gerbing and Anderson, 1992) - that is, χ2/df, RMR, CFI, NFI, and other criteria should have measurement consistency (Steenkamp and Baumgartner, 1998). However, we have to admit that there is still room for improvement in the fitness of this study’s model, which is one of its limitations.

We now examine whether or not the empirical results support the hypotheses. Table 4 displays the structural model with coefficients and the significant relationships between the variables; almost all the variables follow the hypothesized direction. These results provide reasonable evidence to support the study’s model. The endorsement behavior in the structural equation model positively affects information scent (H1a: β1a = 0.016, t-value = 0.117, p-value = 0.224), and hence we reject hypothesis H1a, because the p values are not significant; otherwise, endorsement behavior in the structural equation model has positive effect on message framing (H1b: β1b = 0.168, t-value= 1.758, p-value < 0.05). Sponsorship disclosure has a positive effect on both information scent (H2a: β2a = 0.653, t-value = 4.463, p-value < 0.01) and message framing (H2b: β22 = 0.479, t-value= 4.106, p-value < 0.01). Information scent has a positive effect on stickiness (H3a: β3a = 0.440, t-value = 7.206, p-value < 0.01), and message framing has a positive effect on stickiness (H3b: β3b = .372, t-value = 5.373, p-value < 0.01). Next, the hypothesis is that information scent has a positive effect on message framing (H4: β4 = 0.122, t-value = 2.434, p-value < 0.01). Finally, the last hypothesis is that stickiness has a positive effect on e-WOM (H5: β5 = 0.989, t-value = 1.511, p-value < 0.01).

The effective step for detecting multicollinearity is to examine the correlation among the independent variables. We conduct this by checking a correlation matrix. The correlation between “stickiness” (ST) and “e-WOM”

(17)

Table 4

Test of measurement model

Hypothesized Path Coefficient t p Result H1a: Endorsement Behavior→Information Scent

H1b: Endorsement Behavior→Message Framing H2a: Sponsorship Disclosure→Information Scent H2b: Sponsorship Disclosure→Message Framing H3a: Information Scent→Stickiness

H3b: Message Framing→Stickiness H4: Information Scent→Message Framing H5: Stickiness→ e-WOM 0.016 0.168 0.653 0.479 0.440 0.372 0.122 0.989 0.117 1.758 4.463 4.106 7.206 5.373 2.434 10.511 0.224 0.040 0.000 0.000 0.000 0.000 0.000 0.000 Not supported Supported Supported Supported Supported Supported Supported Supported

(EW) is positive (0.579), while the correlations of the dependent variable (EW) with each of the other two independent variables (“information scent” (IS) and “message framing” (MF)) are also positive with the dependent variables (e-WOM) (0.500 and 0.430). When writing these in shorthand, we use the Greek letter rho, ρ, to denote the correlation. Hence, we express the correlation between each independent variable with the dependent variable as follows: ρ(IS)(EW) = 0.500, ρ(MF)(EW) = 0.430, and ρ(ST)(EW) = 0.579. We now can look at the correlations among our independent variables: ρ(IS)(MF)= 0.431, ρ(IS)(ST)= 0.458, and ρ(MF)(ST)= 0.324. As for the remaining independent variables (CE and SD), the correlation coefficient test is the same and not repeated. We note that all independent variables are not highly correlated with one another (correlation coefficients are all smaller than 0.500). The relationship between “IS” and “MF” is not as strong as the correlation between those individual variables with “ST”, and the correlation is not high. Hence, we cannot conclude that multicollinearity is present in this model (see Table 5).

The mediation effect is usually based on the three-stage analysis proposed by Baron and Kenny (1986). Zhao, Lynch, and Chen (2010) later argued and amended the Baron and Kenny (1986) approach. Additionally, Efron (1979) proposed the bootstrap method, in which the number of existing samples is amplified by repeated sampling, so that the distribution of their number of times is closer to the method of population distribution. With the computer application

(18)

Table 5

Multi-trait multi-method matrix (correlation matrix)

Notes:

1. CE1, CE2, and CE3 are constructs for celebrity endorsement (CD); SD1, SD2, and SD3 are constructs

for sponsorship disclosure (SD); IS1 and IS2 are constructs for information scent (IS); MF1, MF2, MF3 and MF4 are constructs for message framing (MF); ST1, ST2 and ST3 are constructs for stickiness (ST); EW1, EW2, and EW3 are constructs for e-word-of-mouth (EW); AVE is average variance extracted.

2. The values in each column are the average of the correlation coefficients of individual constructs. For

example, the correlation coefficient of celebrity endorsement and celebrity endorsement is 0.789, which is derived from the correlation coefficient of CE1 to CE2 (0.845), the correlation coefficient of CE1 to CE3 (0.722), the correlation coefficient of CE2 to CE3 (0.800). Then, the average of a total of three correlation coefficient values is calculated as 0.789. The correlation coefficient of celebrity endorsement and sponsorship disclosure is 0.551, which is derived from the correlation coefficient of CE1 to SD1 (0.643), the correlation coefficient of CE1 to SD2 (0.613), The correlation coefficient of CE1 to SD3 (0.559), the correlation coefficient of CE2 to SD1 (0.529), the correlation coefficient of CE2 to SD2 (0.541), the correlation coefficient of CE2 to SD3 (0.531), the correlation coefficient of CE3 to SD1 (0.521), The correlation coefficient of CE3 to SD2 (0.536), the correlation coefficient of PE3 to SD3 (0.482). We then compute the average of a total of nine correlation coefficient values and obtain 0.551. The rest of the other cases are similar.

of the structural equation model, the bootstrap method becomes the latest famous method to explore the mediation effect. Next, the bootstrap method is carried out in a virtual manner, and the confidence interval can be calculated via repeated extractions of 1000 times (Efron and Tibshirani, 1993). In the SEM model, because we want to detect and analyze the mediation effect, we use the bootstrap method to obtain the confidence interval of the indirect effect, which can be called the mediation effect if the 95% confidence interval does not contain zero

Variables/constructs (CE) CE1 CE2 CE3 (SD) SD1 SD2 SD3 (IS) IS1 IS2 (MF) MF1 MF2 MF3 MF4 (ST) ST1 ST2 ST3 (EW) EW1 EW2 EW3 AVE CE (CE1、CE2、CE3) 0.789 0.783 SD (SD1、SD2、SD3) 0.551 0.603 0.750 IS (IS1、IS2) 0.396 0.421 0.884 0.824 MF (MF1、MF2、MF3、MF4) 0.504 0.513 0.431 0.708 0.840 ST (ST1、ST2、ST3) 0.369 0.381 0.458 0.324 0.724 0.736 EW (EW1、EW2、EW3) 0.431 0.416 0.500 0.430 0.579 0.720 0.800

(19)

and reaches a significant level. If the direct effect in the 95% confidence interval contains 0, then it means that the direct effect is not significant, with a full mediation effect. If the indirect effect and the direct effect in the 95% confidence interval are not included in 0, and all reach a significant level, then the total effect in the 95% confidence interval does not contain 0, reaching a significant level, and it can be deemed as a partial mediation effect (Mackinnon, 2008; Preacher and Hayes, 2008).

We examine the mediation effects through the bootstrap method from CE (SD) to ST of the study. There are four types of the serial mediation effect: 1) Does MF own a mediation effect between CE and ST (Case 1)? 2) Does IS have a mediation effect between CE and ST (Case 2)? 3) Does MF have a mediation effect between SD and ST (Case 3)? 4) Does IS own a mediation effect between SD and ST (Case 4)? Table 6 lists the results. We find that mediation effects are present in case 1, case 2, and case 4. For example, case 1 shows that the confidence interval (0.034~0.283) of the indirect effect (0.140; 0.140 = 0.634 x 0.221) does not contain 0 and has a significant effect (p < 0.05), indicating that MF has a mediation effect between CE and ST. The direct effect of CE  ST (0.365) of the confidence interval (0.192~0.480) does not contain 0, up to a significant effect, and the total effect (0.505) = direct effect + indirect effect (0.505 = 0.365 + 0.140) of the confidence interval (0.320~0.691) does not contain 0, thus achieving significant results. This shows that the message frame (MF) is a partial mediation effect between celebrity endorsement (CE) and stickiness (ST).

We also examine the mediation effects from the bootstrap method from MF (IS) to EW of the study. There are two types of the serial mediation effect: 1) Does ST own a mediation effect between MF and EW (Case 5)? Does ST have a mediation effect between IS and EW (Case 6)? Table 6 lists the results. We find that they exhibit mediation effects in case 5 and case 6. For example, case 5 shows that the confidence interval (0.178~0.439) of the indirect effect (0.298) does not contain 0 and has a significant effect (p < 0.05), indicating that ST has a mediation effect between MF and EW. The direct effect of MF  ST (0.482) of the confidence interval (0.281~0.664) does not contain 0 up to a significant effect, and similarly the direct effect of ST  EW (0.618) of the confidence interval (0.521~0.733) does not contain 0 up to a significant effect; the total

(20)

Table 6

Empirical result of serial mediation effects

Effects Contents Estimate p-value Confidence interval Case 1: CE  MF  ST Indirect effect CE  MF  ST 0.140 p < 0.05 0.034 ~ 0.283 Direct effect CE  MF MF  ST CE  ST 0.634 0.221 0.365 p < 0.05 p < 0.05 p < 0.05 0.489 ~ 0.766 0.050 ~ 0.391 0.192 ~ 0.480 Total effect CE  ST 0.505 p < 0.05 0.320 – 0.691 Case 2: CE  IS  ST Indirect effect CE  IS  ST 0.208 p < 0.05 0.117 ~ 0.322 Direct effect CE  IS IS  ST CE  ST 0.474 0.439 0.265 p < 0.05 p < 0.05 p < 0.05 0.324 ~ 0.634 0.336 ~ 0.549 0.178 ~ 0.376 Total effect CE  ST 0.473 p < 0.05 0.359 ~ 0.668 Case 3: SD  MF  ST Indirect effect SD  MF  ST 0.010 p < 0.05 -0.155 ~ 0.172 Direct effect SD  MF MF  ST SD  ST 0.752 0.013 0.562 p < 0.05 p < 0.05 p < 0.05 0.623 ~ 0.854 -0.203 ~ 0.231 0.340 ~ 0.833 Total effect SD  ST 0.572 p < 0.05 0.385 – 0.720 Case 4: SD  IS  ST Indirect effect SD  IS  ST 0.203 p < 0.05 0.102 ~ 0.323 Direct effect SD  IS IS  ST SD  ST 0.597 0.391 0.377 p < 0.05 p < 0.05 p < 0.05 0.388 ~ 0.744 0.191 ~ 0.481 0.231 ~ 0.537 Total effect SD  ST 0.581 p < 0.05 0.444 ~ 0.821 Case 5: MF  ST  EW Indirect effect MF  ST EW 0.298 p < 0.05 0.178 ~ 0.439 Direct effect MF  ST ST  EW MF  EW 0.482 0.618 0.340 p < 0.05 p < 0.05 p < 0.05 0.281 ~ 0.664 0.521 ~ 0.733 0.254 ~ 0.433 Total effect MF  EW 0.638 p < 0.05 0.470 – 0.815 Case 6: IS  ST  EW Indirect effect IS  ST EW 0.354 p < 0.05 0.247 ~ 0.482 Direct effect IS  ST ST  EW IS  EW 0.607 0.584 0.303 p < 0.05 p < 0.05 p < 0.05 0.455 ~ 0.744 0.452 ~ 0.745 0.166 ~ 0.425 Total effect IS  EW 0.658 p < 0.05 0.524 ~ 0.811

Notes: CE is celebrity endorsement; SD is sponsorship disclosure; MF is message frames; IS represents information scent; ST is stickiness; EW is e-word-of-mouth.

(21)

effect (0.638) = direct effect + indirect effect (0.638 = 0.340 + 0.298) of the confidence interval (0.470~0.815) does not contain 0, reaching significant results. This shows that the message frame (ST) has a partial mediation effect between message frame (MF) and electronic word-of-mouth (EW).

Bagozzi and Yi (1988) argued for the necessity of a rival model and recommended that one adopt the difference in the χ2 value to compare and examine the validity of the rival model. Kline (2011) employed the values of GFI, CFI, RMSEA, and RMSR and the related indices of the path coefficient to evaluate the model. Kline (2011) also suggested that the significant ratio is the effective decision rule for comparing the study model and the rival models.

The rival model helps check if the causality among mediators is really needed, which originates from the comprehensive school framework (Gefen and Straub, 2004; Lu et al., 2010). Rival model 1 specifies the direct effects between endorsement and sponsorship and the cognitive information effect of website cues without mediators, which are based on the structural school framework (Rapp et al. 2013). Rival model 2 examines if the causality among mediators is really needed, which originates from the comprehensive school framework (Gefen and Straub 2004; Lu et al. 2010). Table 7 helps us analyze the rival models. The direct effect from endorsement and sponsorship to e-loyalty is not more effective than the indirect effect through the mediators, i.e. message framing (MF) and information scent (IS), in the original model. The most important result is that the significant ratio of the original model (87.50%) is larger than rival model 1 (60%) and rival model 2 (50%). Because the original model enjoys the highest significant ratio and has a higher path total effect, we conclude that the original model is better than the rival models. The causal relationship among those variables holds.

5. Discussion and conclusion

One of the most important obstacles to overcome when developing a new website is creating a design that will both attract and stick consumers to the site. Whereas consumers visit hundreds of webpages, they generally only stick or stay on a site for a short time before opting out and moving on to a new site. This paper examines the causal relationship among information disclosure, the

(22)

Table 7

Comparison of rival models

Rival Model 1 Rival Model 2

Significant Ratio 60.00% 50.00%

EB→ST Not Supported EB→IS Not Supported

SD→ST Supported SD→IS Supported

IS→ST Supported EB→MF Not Supported

MF→ST Not Supported SD→MF Not Supported

ST→e-WOM Supported IS→ST Supported

MF→ST Not Supported IS→MF Not Supported

EB→ST Supported SD→ST Supported ST→eWOM Supported χ2/df 5.988 3.660 CFI 0.864 0.930 NFI 0.842 0.907 GFI 0.792 0.864 AGFI 0.724 0.814 RMSEA 0.125 0.091 RMSR 0.134 0.036 Exa m in in g th e E ffec t o f In fo rm atio n D isclo su re o n W eb site Stic kin es s: The Co gni tiv e Inf or m at ion Pe rsp ec tiv e e-WOM Website Stickiness Information Scent Message Framing Sponsorship Disclosure Endorsement Disclosure e-WOM Website Stickiness Information Scent Message Framing Sponsorship Disclosure Endorsement Disclosure

(23)

cognitive information effect of cues, and website stickiness. The cognitive information effect of a website includes its information scent, message framing, information retrieval, and information filtering (Sorensen et al., 1999). Amongst them, information scent and message framing utilize individuals’ short-term memory, and information retrieval and filtering utilize individuals’ long-term memory (Pirolli and Card, 1999). In some emerging industries, the life cycle of products is quite short term. Therefore, individuals’ short-term memory must be highlighted. From the academic viewpoint, while other researchers have focused on information retrieval and filtering to highlight the extraction process in consumers’ long-term memory (Moody and Galletta, 2015), this study emphasize the short-term memory of individuals, i.e. information scent and message framing, to uphold the theoretical contributions in the CIP concept. We combine the message framing and information scent concepts proposed by Rothman (2005) and Gibson et al. (2016) to form the memory mechanism based on the learning process in CIP. In sum, our findings show that sponsorship disclosure, information scent, and message framing are important factors in increasing the stickiness of customers to websites. Both sponsorship disclosure and endorsement behavior confirm the CIP concept and offer further theoretical contributions by applying information scent and message framing.

From the practical viewpoint, through empirical investigations we see that sponsorship disclosure is a critical factor to information scent, which further has a significantly positive effect on stickiness of customers. There may exist two reasons for this. First, sponsorship disclosure provides an information cue to help consumers find a linkage with a positive brand image (Kerne and Smith, 2004). Second, the more the advertising messages align with a consumer’s self-interest, the stronger the persuasion becomes. If purchasing products is expected, then the products purchased are comparable to the consumers’ subjective expectations and perceptions. When consumers’ perception of a product is consistent with their expectation, then their expectation and interest level will be affected by the product information (Viglia et al, 2016). Sponsorship disclosure then is positively related to information scent, while information scent is positively related to stickiness. Moreover, message framing is an important intermediary variable. Sponsorship disclosure also has a positive effect on stickiness through message framing. Sponsorship gives consumers opportunities to find product

(24)

information and allows them to form stickiness. This finding is similar to that of Nebenzahl and Jaffe (1988), in which sponsorship disclosure has a significant impact on consumers’ purchases. Therefore, using sponsorship disclosure as a marketing strategy is definitively imperative (Boerman et al., 2014b; 2017) to managers.

Marketing managers can use sponsorship disclosure or celebrity endorsements for their marketing plans. The advertisements can be used during the product growth period, while the company is able to pay for endorsement and sponsorship due to sufficient profits (Moody and Galletta, 2015). Our study can be used for new product development in both emerging and traditional markets. In fact, brand extension can be increased when celebrities and sponsors are carefully selected during product development. Managers can highlight the importance of finding sponsors who are willing to disclose complete information and ask for cues for the corporate image, reputation, and brand equity of large enterprises (Kaplan and Haenlein, 2010). After applying the solutions proposed by this study, businesses should gain more effective and efficient product development, providing consumers with better information transparency around products (Overton, 2018).

The empirical contribution of this study is to help businesses understand the effects on online consumer behavior, i.e. stickiness and e-WOM, by building consumers’ cognition, i.e. information scent and message framing. This finding is particularly applicable to start-ups in emerging industries such as the online game industry so as to help develop consumers’ stickiness and e-WOM. For start-ups in emerging industries, there are not enough loyal customers to create e-WOM to gradually produce long-term marketing effects. The life cycle of on-line game products is quite short term - namely, it is not easy to form e-WOM through long-term cognition (Hwang and Jeong, 2016). Therefore, based on the findings of this study, start-up companies can use endorsement disclosure and sponsorship disclosure to create e-WOM. This result is similar to that of Hwang and Jeong (2016). Emerging businesses can employ information disclosure so that consumers will be able to recognize their own needs and stimulate themselves to feel the necessary to actions.

Since we find that information scent is the most important factor in stickiness, emerging companies need to design a more convenient and

(25)

frequently-updated website interface to attract consumers on an ongoing basis, as well as to attract others to stay and become sticky consumers. This finding is similar to that of Moody and Galletta (2015). Furthermore, online consumers regard message framing as the resultant message causing their positive attitude (Overton, 2018). Through the operation of information scent and message framing effects, consumers can enjoy the atmosphere of product image and subsequently be interested in it (Mudambi and Schuff, 2010). This is the CIP process of consumers for setting up the latter behavior of stickiness and e-WOM. Therefore, start-up companies do not need to use information retrieval and filtering on the Internet (Weisfeld-Spolter, 2014), but instead can utilize the cognitive effect, i.e. information scent and message framing, to obtain stickiness and e-WOM behavior. Emerging companies should design their website’s information cues for products so as to contribute to information transparency to their consumers (Chiu et al., 2009). The result of this study provides online emerging companies with useful evidence for constructing their relationship with consumers. Comparing the impacts of endorsements with sponsorship disclosure, we find that consumers are more interested in sponsorship disclosure. Thus, emerging companies should design their exposed information about products and website cues to create information transparency with their consumers (Rapp et

al., 2013).

Finally, common method variance is a research limitation. Podsakoff and Organ (1986) indicated that common method variance (CMV) when self-reported surveys are employed as a measurement tool is one potential issue in a behavioral study. The respondents rated their perception of the predictor variable and criterion variable, and the exogenous variables and endogenous variables were collected from the same rater or source (Podsakoff et al., 2003). We present a temporal separation in our online survey by introducing a cover picture and short story between the independent variable and dependent variable so as to create a time lag to ensure that the measurement of the independent variable is not directly connected to the dependent variable (Podsakoff et al., 2003). Our research design (temporal separation) tries to handle the issue of common method variance by decreasing the perceived relevance of previously recalled information in short-term memory. Using Harman’s one-factor analysis of unrotated principal components method, we see that there are nine factors

(26)

with eigenvalues greater than 1.0 rather than a single factor within the 54 items. The nine factors together account for 76.10% of the total variance, and no general factors are obvious. Additionally, the interpretation variation of the first factor is 41.96%, meaning that most of the explanatory power is not explained by the first factor, showing that there is no CMV in this measurement questionnaire and that CMV issue may not be a serious problem in our study (Malhotra, Kim and Patil 2006). We do admit that there is still room for improvement in the survey process. This can be deemed as a limitation of this study and could be improved by future efforts.

This study aims to investigate stickiness through the cognitive information effect of website cues. Therefore, our questionnaire focuses on the answers from website consumers instead of those from website designers. The dyad samples can collect more effective information and can be set as the next target in the future. Moreover, length of time and frequency and depth of visit, rather than stickiness only, are the ultimate goals of the online emerging platform marketers. Therefore, transforming stickiness into these goals can be another issue for online emerging platform marketers to evaluate and analyze. In the future, research could pay more attention to this issue in order to provide more useful strategies for managers.

References

Anderson, E. and Barton, W. (1992). The use of pledges to build and sustain commitment in distribution channels. Journal of Marketing Research, 29(2), 18-34.

Anderson, J. C. and Gerbing, D.W. (1988). Structural equation modeling in practice: A review and recommends two-step approach. Psychological

Bulletin, 103(3), 311-323.

Bagozzi, R. P. and Yi, Y. (1988). On the evaluation of structure equations models. Academic of Marketing Science, 16(1), 76-94.

Baron, R. M. and Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology,

(27)

Bickart, B. and Schindler, R. M. (2001). Internet forums as influential sources of consumer information. Journal of Interactive Marketing, 15(3), 31-40.

Boerman, S., Reijmersdal, E. A., and Neijens, P. C. (2012). Sponsorship disclosure: Effects of duration on persuasion knowledge and brand responses. Journal of Communication, 62(6), 1047-1064.

Boerman, S., Reijmersdal, E. A., and Neijens, P. C. (2014a). Effects of sponsorship disclosure timing on the processing of sponsored content: A study on the effectiveness of European disclosure regulations. Psychology &

Marketing, 31(3), 214-224.

Boerman, S., Reijmersdal, E. A., and Neijens, P. C. (2014b). How audience and disclosure characteristics influence memory of sponsorship disclosures.

International Journal of Advertising, 34(4), 576-592.

Boerman, S. C., Willemsen, L. M., and VanDer, E. P. (2017). This post is sponsored: Effects of sponsorship disclosure on persuasion knowledge and electronic word of mouth in the context of facebook. Journal of Interactive

Marketing, 38(1), 82-92.

Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.

Boyd, M. D. and Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13(1), 210-230.

Chen, C. Y., Huang, H. H., and Wey, S. C. (2017). The mediating roles of differentiation strategy and learning orientation in the relationship between entrepreneurial orientation and firm performance. Corporate Management

Review, 37(1), 1-40.

Cheng, H. W., Chien, C. L., and Woodburne, P. R. (2017). The impact of cause-related marketing on store switching: An analysis of the chain convenience stores in Taiwan. Corporate Management Review, 37(2), 87-126.

Chiu, C. M., Chang, C. C., Cheng, H. L., and Fang, Y. H. (2009). Determinants of customer repurchase intention in online shopping. Online Information

Review, 33(4), 761-784.

Chou, C. K. and Chen, M. L. (2016). A qualitative study on perceived value and loyalty: A moderated-mediation framework. Corporate Management Review,

(28)

36(2), 105-122.

Chong, D. and Druckman, J. N. (2007). A theory of framing and opinion formation in competitive elite environments. Journal of Communication,

57(3), 99-118,

Churchill, G. A. and Surprenant, C. (1982). An investigation into the determinants of customer satisfaction. Journal of Marketing Research,

19(11), 491-504.

Dodds, W. B., Monroe, K. B., and Grewall, D. (1991). Effects on price, brand and store information on buyers’ product evaluations. Journal of Marketing

Research, 28(8), 307-319.

Efron, B. (1979). Bootstrap methods: Another look at the Jackknife. The Annals

of Statistics, 7(1), 1-26.

Efron, B. and Tibshirani, R. (1993). An introduction to the bootstrap. New York: Chapman & Hall/CRC.

Fleck, N., Korchia, M., and Roy, I. L. (2012). Celebrities in advertising: looking for congruence or likability? Psychology and Marketing, 29(9), 651-662. Fornell, C. and Larcker, D. F. (1981). Evaluating structural equation models with

unobservable variables and measurement error. Journal of Marketing

Research, 18(1), 39-50.

Gefen, D. and Straub, D. W. (2004). Consumer trust in B2C e-commerce and the importance of social presence: Experiments in e-products and e-services.

Omega, 32(6), 407-424.

Gerbing, D. W. and Anderson, J. C. (1992). Monte Carlo evaluations of goodness of fit indices for structural equation models. Sociological Methods

& Research, 21(2), 132-160.

Gibson, B., Butler, J., Zirkle, M., Hammond, K., and Weir, C. (2016). Foraging for information in the EHR: The search for adherence related information by mental health clinicians. AMIA Annual Symposium Proceedings, 600-608. Greenwald, A. (1968). Cognitive learning, cognitive response to persuasion, and

attitude change. In Greenwald, A., Brock, T. and Ostrom, T. (Eds),

Psychological foundations of attitudes (pp. 147-170), New York, NY:

Academic Press.

Han, C. F. (1988). The role of consumer patriotism in the choice of domestic versus foreign products. Journal of Advertising Research, 28(3), 25-35.

(29)

Hennig-Thurau, T., Gwinner, K. P., Walsh, G., and Gremler, D. D. (2004). Electronic word-of-mouth via consumer-opinion platforms: What motivates consumers to articulate themselves on the internet? Journal of Interactive

Marketing, 18(11), 38-52.

Ho, S. Y. and Bodoff, D. (2014) The effects of web personalization on user attitude and behavior: An integration of the elaboration likelihood model and consumer search theory, MIS Quarterly, 38(2), 497-520.

Holland, J. and Baker, S. M. (2001). Customer participation in creating site brand loyalty. Journal of Interactive Marketing, 15(4), 34-45.

Hsu, F. M., Fan, C. T., Chen, T. Y., and Wang, S. W. (2013). Exploring perceived value and usage of information systems in government context. Chiao Da

Management Review, 33(2), 75-104.

Hu, L. T. and Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternative. Structural

Equation Modeling, 6(1), 1-55.

Hwang, Y. and Jeong, S. H. (2016). “This is a sponsored blog post, but all opinions are my own”: The effects of sponsorship disclosure on responses to sponsored blog posts. Computers in Human Behavior, 62(4), 528-535. Johar, J. S. and Sirgy, J. M. (1991). Value-expressive versus utilitarian

advertising appeals: When and why to use which appeal. Journal of

Advertising, 20(1), 23-33.

Jöreskog, K. G. and Sörbom, D. (1993). LISREL 8: Structural equation modeling

with simplis command language. Mooresville, Chicago, IL: Scientific

Software International, Inc.

Kamins, M. A., Brand, M. J., Hoeke, S. A., and Moe, J. C. (1989). Two-sided versus one-sided celebrity endorsement: The impact on advertising effectiveness and credibility. Journal of Advertising, 18(2), 4-10.

Kaplan, A. M. and Haenlein, M. (2010). Users of the world, unite: The challenges and opportunities of social media. Business Horizons, 53(1), 59-68.

Keng, C. J., Chen, Y. H., and Huang, Y. H. (2018). The influence of mere virtual presence with product experience and social virtual product experience on brand attitude and purchase intention: Conformity and social ties as moderators. Corporate Management Review, 38(2), 57-94.

(30)

Kline, R. B. (2011). Principles and practices of structural equation modeling (3rd ed.). New York, NY: Guilford Publications.

Kwon, O. and Sung, Y. J. (2012). Shifting selves and product reviews: How the effects of product reviews vary depending on the self-views and self-regulatory goals of consumers. International Journal of Electronic

Commerce, 17(1), 59-82.

Kurstedt, H. A. Jr. (1987). Responsive decision support systems: A broad view illustrates when to include expert systems. In H. G. Sol, C. A. Th. Takkenberg, and P. F. de Vries Robbé (Eds.), Expert systems and artificial

intelligence in decision support Systems. (pp. 53-78.). Dordrecht, Holland: D.

Reidel Publishing Company.

LaBerge, D. and Samuels, J. S. (1974). Toward a theory of automatic information processing in reading. Cognitive Psychology, 6(2), 293-323. Lin, J. C. C. (2007). Online stickiness: Its antecedents and effect on purchasing

intention. Behavior and Information Technology, 26(6), 507-516.

Lu, Y. B., Zhao, L., and Wang, B. (2010). From virtual community members to C2C e-commerce buyers: trust in virtual communities and its effect on consumers’ purchase intention. Electronic Commerce Research and

Applications, 9(4), 346-360.

MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. New York, NY: Lawrence Erlbaum Associates.

Malhotra, N. K., Kim, S. S., and Patil, A. (2006). Common method variance in IS research: A comparison of alternative approaches and a reanalysis of past research. Management Science, 52(12), 1865-1883.

McCracken, G. (1989). Who is the celebrity endorser? Cultural foundation of the endorsement process. Journal of Consumer Research, 16(3), 310-321. McIntyre, N. and Pigram, J. J. (1992). Recreation specialization reexamined: The

case of vehicle-base campers. Leisure Sciences, 14(1), 3-15.

Meyers-Levy, J. and Maheswaran, D. (2004). Exploring message framing outcomes when systematic, heuristic or both types of processing occur.

Journal of Consumer Psychology, 14(1), 159-167.

Moody, G. D. and Galletta, D. F. (2015). Lost in cyberspace: The impact of information scent and time constraints on stress, performance, and attitudes online. Journal of Management Information Systems, 32(1), 192-224.

(31)

Mudambi, S. M. and Schuff, D. (2010). What makes a helpful online review? A study of customer review on amazon.com. MIS Quarterly, 34(1), 185-200. Nebenzahl, I. D. and Jaffe, E. D. (1998). Ethical dimensions of advertising

executions. Journal of Business Ethics, 17(7), 805-815.

Nelson, M. R., Wood, M. L., and Paek, H. J. (2009). Increased persuasion knowledge of video news releases: Audience beliefs about news and support for source disclosure. Journal of Mass Media Ethics, 24(2), 220-237.

Nisbet, M. C. (2009). Knowledge into action: Framing the debates over climate change and poverty. In P. D’Angelo and J. Kuypers (Eds.), Doing news

framing analysis: Empirical, theoretical, and normative perspectives (pp.

46-83). New York, NY: Routledge.

Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York. NY: McGraw-Hill.

Ohanian, R. (1991). The impact of celebrity spokespersons' perceived image on consumers' intention to purchase. Journal of Advertising Research, 31(1), 46-54.

Overton, H. K. (2018). Examining the impact of message frames on information seeking and processing: A new integrated theoretical model. Journal of

Communication Management, 22(3), 327-345.

Pfeffer, J. (1981). Management as symbolic action: The creation and maintenance of organizational paradigms. In L. L. Cummings and B. Staw (Eds.), Research in Organizational Behavior (Vol. 3). Greenwich, CT: JAI Press.

Pirolli, P. and Card, S. K. (1999). Information foraging. Psychological Review,

106(4), 643-675.

Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., and Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(6), 879-891.

Podsakoff, P. M. and Organ, D. W. (1986). Self-reports in organizational research: Problems and prospects. Journal of Management, 12(4), 531-544. Preacher, K. J. and Hayes, A. F. (2008). Asymptotic and resampling strategies for

assessing and comparing indirect effects in multiple mediator models.

(32)

Rajala, A. K. and Hantula, D. A. (2000). Towards a behavioral ecology of consumption: Delay-reduction effects on foraging in a simulated internet mall. Managerial and Decision Economics, 21(1), 145-158.

Rapp, A., Beitelspacher, L. S., Grewal, D., and Hughes, D. E. (2013). Understanding social media effects across seller, retailer, and consumer interactions. Journal of the Academy of Marketing Science, 41(5), 547-566. Richardson, P., Dick, A., and Jain, A. (1994). Extrinsic and intrinsic cue effects

on perceptions of store brand quality. Journal of Marketing, 58(4), 28-36. Ross, I. (1979). Reviewed work: An information processing theory of consumer

choice by James R. Bettman. Journal of Marketing, 43(1), 124-126.

Rothman, L. S. (2005). The HITRAN 2004 molecular spectroscopic database source. Journal of Quantitative Spectroscopy & Radiative Transfer, 96(2), 139-204.

Sandler, D. M. and Shani, D. (1993). Sponsorship and the Olympic game: The consumer perspective. Sport Marketing Quarterly, 2(3), 38-44.

Sorensen, H., O'Riordan, A., and O'Riordan, C. (1999). Profiling with the INFOrmer text filtering agent. Journal of Universal Computer Science, 3(8), 988-1006.

Spool, J. M., Perfetti, C., and Brittan, D. (2004). The essentials every designer

needs to know about howvusers navigate through large web sites, Middleton,

MA: User Interface Engineering.

Steenkamp, J. E. M. and Baumgartner, H. (1998). On the use of structural equation models for marketing modeling. International Journal of Research

in Marketing, 17(2), 195-202.

Stipp, H. and Schiavone, N. P. (1996). Modeling the impact of Olympic sponsorship on corporate image. Journal of Advertising Research, 36(4), 22-28.

Taiwan Network Information Center (2017). Taiwan broadband network usage

survey. Taiwan Network Information Center (TWNIC), Taipei (in Chinese).

Viglia, G., Minazzi, R., and Buhalis, D. (2016). The influence of e-word-of-mouth on hotel performance: Investigating the determinants of occupancy rate. International Journal of Contemporary Hospitality

Management, 28(9), 2035-2051.

(33)

you get what you pay for? Journal of Consumer Marketing, 9(2), 69-76. Weisfeld-Spolter, S. (2014). An integrative approach to e-wom and marketing

communications. Corporate Communications: An International Journal,

19(3), 260-274.

Wu, C. M. (2011). What factors affect people's stickiness on the social network site: the case of Facebook in Taiwan. Key Engineering Materials, 474-476, 1573-1577.

Zhao, X., Lynch Jr, J. G., and Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and truths about mediation analysis. Journal of Consumer Research,

37(2), 197-206.

Zott, C., Amit, R. B., and Donlevy, J. (2000). Strategies for value creation in e-commerce: best practice in Europe. European Management Journal, 18(5), 463-475.

數據

Figure 1    Research framework  e-WOM Information Scent Website Stickiness Endorsement Disclosure Sponsorship Disclosure Message Framing Information Disclosure Cognitive Information Effect E-loyalty
Table 1    Sample structure  Items  18-49  years-old  50-64

參考文獻

相關文件

An information literate person is able to recognise that information processing skills and freedom of information access are pivotal to sustaining the development of a

Disclosure and Transparency: The corporate governance framework should ensure that timely and accurate disclosure is made on all material matters regarding the corporation,

ix If more than one computer room is opened, please add up the opening hours for each room per week. duties may include planning of IT infrastructure, procurement of

Adolescents who conducted doxing had greater odds of disclosing others’ personal information, students who had conducted doxing had also experienced information disclosure as

Adolescents who conducted doxing had greater odds of disclosing others’ personal information, students who had conducted doxing had also experienced information disclosure as

„ Indicate the type and format of information included in the message body. „ Content-Length: the length of the message

 The IEC endeavours to ensure that the information contained in this presentation is accurate as of the date of its presentation, but the information is provided on an

• The unauthorized disclosure of information could be expected to have a severe or catastrophic adverse effect on organizational operations, organizational assets, or