CHAPTER 2: LITERATURE REVIEW
2.8 Herd Behavior
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
2.8 Herd Behavior
Some online platforms are incorporating website features to verify whether product reviews or comments are generated by real consumers (DeAndrea et al., 2015). Expedia.com only allows customers with confirmed transactions to post product reviews on their website.
Despite Facebook has not yet introduced similar verification features (Corporate Facebook pages are open to all Facebook users), the functions of “Like”, “Share” and “Comment” on Facebook allow the viewers to comprehend the opinions of other consumers. On Facebook, electronic word-of-mouth (eWOM) of a brand can be spread by users liking, sharing and commenting on the posts (Hsu & Liu, 2013), while “Like” is the function most commonly used by users to show that they agree, support, like, or are interested in the content.
H3a: On corporate Facebook Pages, positive consumer-generated content is perceived to have more positive effect on intention to click “Like” for posts than corporate-generated content.
H3b: On Corporate Facebook Pages, positive identifiable consumer-generated content is perceived to have more positive effect on intention to click “Like” for posts than positive unidentifiable consumer-generated content.
H3c: On Corporate Facebook Pages, corporate-generated content is perceived to have more positive effect on intention to click “Like” for posts than positive unidentifiable consumer-generated content.
When groups of consumers share information or express their opinions about products and services, their attitudes or behavior sometime align without centralized coordination.
This phenomenon is known as herding. Herd behavior describes how individuals in a group
‧
can act collectively without centralized direction. Banerjee (1992) utilized herd behavior in the area of human society and described people’s decision making process from a marketing perspective. Banerjee (1992) investigated the herd behavior in a population of agents who make choices between assets based on their own information and also on the observed behavior of other agents. It showed that decision makers look at the judgments made by previous decision makers in taking their own decisions. In other words, people tend to converge on similar behavior, resulting in a situation with people doing what others are doing rather than using their own information. The decision made is dependent on the population level of the decision and the focus is on the group of people expressing a behavior (Banerjee, 1992). For example, people often decide on what restaurants to visit based on how popular they seem to be.
There are significant differences between traditional (offline) and online herd behavior (Langley, Hoeve, Ortt, Pals, & van der Vecht (2014). According to Barry and Fulmer (2004), three important attributes of the online setting are social bandwidth, interactivity and surveillance. Langley et al. (2014) proposed that these attributes are relevant to herding as they change the way that people interact and influence each other. Social bandwidth, which refers to the transmission of socially relevant information via the online medium, will affect the process of herding by providing social identity cues through which individuals can assess their similarity to others in the herd. For example, on a Corporate Facebook Page, users are devoted to a specific topic. Interactivity also play a role in herding in the online setting, as individuals can interact with a wide range of others both real-time and asynchronously.
Moreover, many-to-many discussions between a large number of people is possible online.
Surveillance refers to the publicly observable nature of much online information. A herd can make its intention visible via an online application such as Corporate Facebook Page, allowing many other users or followers to see what is happening and follow developments, without the need for the intervention of traditional media. Taken together, these distinctive
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
attributes of online setting and social media represent a foundation for a change in herding dynamics (Langley et al., 2014).
Herd behavior in the online setting has received research attention in the marketing and management literature. Peterson and Merino (2003) showed that consumers tend to search for information to reduce uncertainty and risk in the computer-mediated communication (CMC) environment. They make decisions based on existing online information. However, when facing extensive amount of information, people often imitate others rather than making decisions based on existing conditions (Bonabeau, 2004). They tend to follow the previous behavior of others and disregard their own existing information. Salganik, Dodds, and Watts (2006) suggested that consumers influence each other when downloading music from a web application. By comparing the conditions which individuals download songs after listening to them versus which they also see the download choices of others, it was shown that social influence can drive online herd behavior and the herd behavior can produce unpredictable outcomes in terms of which songs become popular within a population.
H4: On Corporate Facebook Pages, corporate-generated content with more “likes” is perceived to have more positive effect on (H4a) corporate reputation, (H4b) purchase intention and (H4c) intention to click “Like” for posts respectively than that with less
“likes”.
H5: On Corporate Facebook Pages, positive identifiable consumer-generated content with more “likes” is perceived to have more positive effect on (H5a) corporate reputation, (H5b) purchase intention and (H5c) intention to click “Like” for posts respectively than that with less “likes”.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
21
H6: On Corporate Facebook Pages, positive unidentifiable consumer-generated content with more “likes” is perceived to have more positive effect on (H6a) corporate reputation, (H6b) purchase intention and (H6c) intention to click “Like” for posts respectively than that with less “likes”.
‧
3.1 Experimental Design Overview
In this study, a 3 (content on Corporate Facebook Page) x 3 (number of “post likes”) between-subject experimental design has been employed. The experiment was conducted online. Participants were randomly assigned to one of the nine conditions. Each of the participants received the website address (URL) of the experiment which connected him/her to a mock-up Corporate Facebook Page. After reading the content on the Corporate Facebook Page, the participants were directed to an online questionnaire which aimed at measuring their perceived corporate reputation of the target, intention to purchase and intention to click
“Like” for the posts.
For the content on Corporate Facebook Page factor, participants were randomly assigned to read one of three posts on a mock-up restaurant Corporate Facebook Page, which are generated by the target corporate, identifiable consumers and unidentifiable consumers respectively. This factor corresponds to H1 and compares the effect of corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content on corporate reputation.
For the number of “post likes” factor, participants were randomly assigned to read one of the three posts with “many likes”, “few likes” and not showing the number of likes. The posts without showing the number of likes serve the purpose of testing the effect of corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content on corporate reputation without being moderated by the number of post likes i.e. herd behavior. Participants read the content solely and determine how they perceived the reputation of the target without being affected by the opinions of others. The posts with “many likes” and “few likes” were thus for testing how herd behavior moderated
‧
the effect of the content on corporate reputation. “Like” is the function most commonly used by Facebook users to show that they agree, support, like, or are interested in the content (Hsu
& Liu, 2013). Therefore, the number of likes represented the number of other Facebook users who agreed with the content.
After viewing the stimulus materials, participants completed an online questionnaire to measure their perceived corporate reputation of the target, intention to purchase and intention to click “Like” for the posts. The questionnaire is based on the “Corporate Reputation Measurement Model” proposed by Han et al. (2015), while the intention to purchase was accessed using the model proposed by Ashton, Scott, Solnet and Breakey (2010) and intention to click the “Like” button was assessed with the model proposed by Chin, Lu and Wu (2015).
Table 1
Experimental design
Content on Corporate Facebook Pages
Corporate-generated
‧
Content on Corporate Facebook Pages Corporate-generated Content
Identifiable Consumer-generated Content Unidentifiable Consumer-generated
Content
Content on Corporate Facebook Pages with different number of “Likes”
Corporate-generated Content with “Many Likes” or “Few Likes”
Identifiable Consumer-generated Content with “Many Likes” or “Few Likes”
Unidentifiable Consumer-generated Content with “Many Likes” or “Few
Likes”
‧
The sample consisted of 270 participants, who were the students and staff members of National Chengchi University and friends of the researcher, recruited by an invitation email which included the website address (URL) of the mock-up Corporate Facebook Page and questionnaire. Participants were randomly assigned to one of the nine groups (9 groups, 30 participants each group) in the main experiment.
Participants ranged in age from 20 years old or below (6.7%) to 51-60 years old (0.7%).
Most participants were 21-30 years old (75.9%). The participants were identified to be residing in Taiwan (53.3 Hong Kong (41.1%), Macau (2.6%), and other places (3%) including Mainland China, Singapore and the USA. More participants identified as female (61.1%) than male (38.9%).
Residence place Taiwan 144 (53.3%)
Hong Kong 111 (41.1%)
Macau 7 (2.6%)
Others (Mainland China, Singapore and the USA) 8 (3%)
‧
A mock-up restaurant Corporate Facebook Page was created for a restaurant that does not exist and used in this study. Prior to the main experiment, a manipulation check was conducted to select the stimulus materials for each group. The manipulation check aimed at selecting: 1) Facebook user profile of identifiable consumers and unidentifiable consumers; 2) posts generated by the restaurant (official corporate information), identifiable consumers and unidentifiable consumers; and 3) posts with “many likes” and “few likes” in the perceptions of participants. For the Facebook user profile of identifiable consumers and unidentifiable consumers, clues included profile pictures and user names. For the generator of the posts, whether they were posted by the restaurant, identifiable or unidentifiable consumers, clues included the content and tone of the information, and photographs uploaded.
The manipulation check had 40 participants, who were excluded from participating in the main experiment. Chi-square test for independence was adopted to test whether there is a significant relationship among the variables of the manipulation check. The stimulus materials with significance level p < 0.05 were taken and adopted in the main experiment.
After the manipulation check, an invitation email which included the website address (URL) of the mock-up Corporate Facebook Page and questionnaire was sent to the participants. The website address (URL) of the experiment connected the participants to a mock-up restaurant Corporate Facebook Page. After reading the content on the Corporate Facebook Page, the participants were then directed to an online questionnaire which aimed at measuring their perceived corporate reputation of the target, intention to purchase and intention to click “Like” for the posts.
The screenshots of the stimulus materials are shown in appendix 2, figure 5 to figure 13.
‧
The perception of corporate reputation was accessed using the “Corporate Reputation Measurement Model” proposed by Han et al. (2015). The model is a 24-item instrument to measure the consumer-based corporate reputation of chain restaurants along 7 dimensions:
food and service quality, brand affect, self-congruence, brand awareness, brand association, brand trust and overall brand reputation. Participants were asked to rate their agreement with the statements on a seven-point Likert-type scale (ranging from 1 = “Strongly disagree” to 7
= “Strongly agree”). The dimensions and the statements are shown below:
1. Food and service quality:
The staff of this restaurant brand is helpful and friendly;
This restaurant brand offers a tidy environment;
This restaurant brand provides comfortable seats and tables;
The staff of this restaurant brand is talented and displays a natural expertise;
This restaurant brand offers fresh foods; and
This restaurant brand prepares food and drinks according to hygiene standard.
2. Brand affect:
I think I will feel good when I dine in this restaurant brand;
I believe this restaurant brand will make me happy; and
I believe this restaurant brand will give me pleasure.
3. Self-congruence:
The customers who dine in this restaurant are very much like me;
The customers who dine in this restaurant reflect the type of person I would like to be; and
The customers who dine in this restaurant are very much like the person I admire.
‧
I am familiar with this restaurant brand; and
I can recognize this brand among other restaurant brands.
5. Brand association:
This brand has an attractive logo;
I like the logo of this brand; and
I like the colors of building or interior of this brand.
6. Brand trust:
I can rely on this brand to solve the service dissatisfaction;
This brand guarantees satisfaction; and
I have confidence in this brand.
7. Overall brand reputation:
This brand is trustworthy;
This brand is reputable; and
This brand makes honest claims.
The intention to purchase was accessed using the model proposed by Ashton et al.
(2010). It included 4 items to measure the intention to purchase in restaurant dining.
Participants were asked to rate their agreement with the statements on a seven-point Likert-type scale (ranging from 1 = “Strongly disagree” to 7 = “Strongly agree”). The statements are shown below:
1. I might dine in the restaurant in the future;
2. I would certainly dine in this restaurant;
3. I would consider dining in this restaurant;
4. I would recommend this restaurant to the others.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
29
The intention to click “Like” for the posts was assessed using the model proposed by Chin et al. (2015). The instrument included 3 items to measure Facebook users’ intention to click the “Like” button for the posts. Participants were asked to rate their agreement with the statements on a seven-point Likert-type scale (ranging from 1 = “Strongly disagree” to 7 =
“Strongly agree”). The statements are shown below:
1. I am very willing to click “Like” after reading this post;
2. I would like to click “Like” for this post;
3. I do not think I should click “Like” for this post.
3.5 Pilot Test
Before testing the hypotheses, a pilot test was conducted to confirm the design of the stimulus materials and the questionnaire. 45 participants (9 groups of stimulus materials, 5 participants in each group) took part in the pilot test. The website address (URL) of the mock-up Corporate Facebook Page and questionnaire was sent to the participants. After viewing the mock-up Facebook Page, they were directed to the online questionnaire to measure their perceived corporate reputation of the restaurant, purchase intention and intention to click “Like” for the posts. The questionnaires were collected and the data was entered into the SPSS. Analysis indicated that the scales for corporate reputation along the 7 dimensions (food and service quality, brand affect, self-congruence, brand awareness, brand association, brand trust and overall brand reputation), purchase intention and intention to click “Like” for posts were reliable (α value > 0.7). The corporate reputation measurement scale was reliable with Cronbach’s α = 0.859, the intention to purchase scale was reliable with α = 0.752, and the intention to click “Like” for posts scale was reliable with α = 0.925.
‧
4.1 Effect of Content on Corporate Reputation
An analysis of variance (ANOVA) was conducted to test Hypothesis 1. It aimed at examining and comparing the effects of corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content on corporate reputation. The corporate reputation variable was created from a 24-item seven-point Likert scale by accumulating the score of each item. The scale was reliable, α = 0.961, n = 270.
According to the results of ANOVA, corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content had significantly different effects on corporate reputation, F(2, 87) = 6.49, df = 2, p < 0.05. Thus, fisher least significant difference (LSD) test was then conducted to further investigate the differences of the effects (figure 2).
Supporting H1a, participants in Group 2 (the identifiable consumer-generated content condition) reported that the identifiable consumer-generated content had significantly more positive effect on corporate reputation, M = 5.28, SD = 1.01, relative to participants in Group 1 (the corporate-generated content condition), M = 4.77, SD = 0.51, F(2, 87) = 6.49, df = 2, p
< 0.05. Next, participants in Group 2 reported that the identifiable consumer-generated content had significantly more positive effect on corporate reputation, M = 5.28, SD = 1.01, relative to participants in Group 3 (the unidentifiable consumer-generated content condition), M = 4.53, SD = 0.89, F(2, 87) = 6.49, df = 2, p < 0.05, supporting H1b.
Regarding the comparison between the effect of corporate-generated content and unidentifiable consumer-generated content on corporate reputation, participants in Group 1 reported that the corporate-generated content had more positive effect on corporate reputation, M = 4.77, SD = 051, than unidentifiable consumer-generated content (participants in Group
‧
statistics, corporate-generated content had stronger effect on corporate reputation than unidentifiable consumer-generated content, yet the effect was not significant according to LSD. Therefore, H1c is not supported.Figure 2. Effects of corporate Facebook page content on corporate reputation.
Note. Group 1 = Corporate-generated content without showing number of “Likes”;
Group 2 = Identifiable consumer-generated content without showing number of “Likes”;
Group 3 = Unidentifiable consumer-generated content without showing number of “Likes”.
4.2 Effect of Content on Intention to Purchase
An analysis of variance was conducted to test Hypothesis 2, examining the effect of corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content on purchase intention. The purchase intention variable was created from a 4-item seven-point Likert scale by accumulating the score of each item. The scale was reliable, α = 0.892, n = 270. According to the results of ANOVA, corporate-generated content, identifiable consumer-generated content and unidentifiable
4.77
Effects of Corporate Facebook Page Content on Corporate Reputation
Group 1 Group 2 Group 3 2
0
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
32
consumer-generated content had significantly different effects on purchase intention, F(2, 87)
= 4.23, df = 2, p < 0.05. Thus, fisher least significant difference (LSD) test was then conducted to further investigate the differences of the effects (figure 3).
Participants in Group 2 reported that the identifiable consumer-generated content had more positive effect on purchase intention, M = 5.40, SD = 1.14, relative to the corporate-generated content in Group 1, M = 5.15, SD = 0.55, F(2, 87) = 4.23, df = 2, p = 0.296 > 0.05. Descriptive analysis shows that identifiable consumer-generated content had stronger effect on intention to purchase than corporate-generated content. However, ANOVA shows the effect to be not significant. H2a is not supported.
Supporting H2b, participants in Group 2 reported that the identifiable consumer-generated content had significantly more positive effect on purchase intention, M = 5.40, SD = 1.14, relative to the unidentifiable consumer-generated content in Group 3, M = 4.72, SD = 0.97, F(2, 87) = 4.23, df = 2, p < 0.05.
H2c is not supported as participants in Group 1 reported that the corporate-generated content had more positive effect on purchase intention, M = 5.15, SD = 0.55, relative to the unidentifiable consumer-generated content in Group 3, M = 4.72, SD = 0.97, F(2, 87) = 4.23, df = 2, p = 0.072 > 0.05. According to descriptive statistics, corporate-generated content had
stronger effect on purchase intention than unidentifiable consumer-generated content, yet the effect was not significant on ANOVA.
‧
Figure 3. Effects of corporate Facebook page content on purchase intention.
Note. Group 1 = Corporate-generated content without showing number of “Likes”;
Group 2 = Identifiable consumer-generated content without showing number of “Likes”;
Group 3 = Unidentifiable consumer-generated content without showing number of “Likes”.
4.3 Effect of Content on Intention to Click “Like” for Posts
An analysis of variance was conducted to test Hypothesis 3, examining and comparing the effect of corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content on intention to click “Like” for the posts. The intention to click “Like” for posts variable was created from a 3-item seven-point Likert scale by accumulating the score of each item. The scale was reliable, α = 0.934, n = 270. According to the results of ANOVA, corporate-generated content, identifiable consumer-generated content and unidentifiable consumer-generated content did not have significantly different effects on intention to click “Like” for posts, F(2, 87) = 1.82, df = 2, p = 0.167 > 0.05. H3 is not supported.
Participants in Group 2 reported that identifiable consumer-generated content had more positive effect on intention to click “Like” for posts (figure 4), M = 4.78, SD = 1.54, than
Effects of Corporate Facebook Page Content on Intention to Purchase
Group 1 Group 2 Group 3 2
0
‧
corporate-generated content reported by participants in Group 1, M = 4.39, SD = 1.22, than
corporate-generated content reported by participants in Group 1, M = 4.39, SD = 1.22, than