an empirical study on identifying opinion leaders in...

Global Journal of Emerging Trends in e-Business, Marketing and Consumer Psychology (GJETeMCP)

An Online International Research Journal (ISSN: 2311-3170)

2018 Vol: 4 Issue: 1

581 www.globalbizresearch.org

An Empirical Study on Identifying Opinion Leaders in the Online

Communities of Consumption

YU WANG,

Department of Economic and Management,

Tohoku University, Japan.

E-mail: [email protected]

___________________________________________________________________________

Abstract

Nowadays, the Internet users begin to search and share electronic word-of-mouth (eWOM)

information online. Meanwhile, they begin to rely on online opinion leaders who can filter the

eWOM information. Considering the importance and influence of online opinion leaders

increasing, the identification of opinion leaders is crucial to marketers and researchers.

Previous researches show that three kinds of approaches, including the user attributes analysis,

the text mining analysis, and the network structure analysis, are widely used for identifying

online opinion leaders. However, litter research is about identifying opinion leaders in the

online communities of consumption, let alone in the online communities of consumption in

which they cannot be identified directly. Hence, this article had one of the first empirical studies

on identifying the opinion leaders in such online communities. The result confirmed the

applicability of selected approach, called Social Network Analysis, for identifying the opinion

leaders in such online communities of consumption, and provided theoretical and practical

implications for marketers and researchers.

___________________________________________________________________________

Key Words: Electronic word-of-mouth (eWOM), Online communities of consumption, Opinion

leaders, Social Network Analysis (SNA)





1. Introduction

Nowadays, Internet users get used to interacting with others and sharing eWOM information

online in various online platforms, such as bulletin boards, discussion forums, chatrooms,

emails, newsgroups, blogs, microblogs and so on. In particular, Internet users with similar

interests or habits gather and interact online, leading to the form and popular of the online

communities. For example, the potential or actual consumers gather in the online communities

of consumption for sharing the information towards products or services (Hagel & Armstrong,

1997).

However, limited by time, it is impossible for Internet users to check all the information.

Hence, they begin to rely on the recommendations from opinion leaders who filter the

information and pass the selected ones to the public. With the number of eWOM information

increasing and the opinion leaders becoming popular, researchers began to notice the influences

of opinion leaders during the process of spreading eWOM (Summers, King, Martin, & Jackson,

1976). These opinion leaders can efficiently convey the marketing messages and meanings to

a wider population (Kozinets, 2010), and can affect the purchase behavior of consumers (Stern

& Gould, 1988; Goldsmith et al., 2003; Rogers, 2003; Valente & Pumpuang, 2007).

Considering the importance and influence of opinion leaders, the companies, marketers and

the online communities begin to utilize opinion leaders as tools for marketing (Stern & Gold,

1988).

Before utilizing opinion leaders, they need to identify the opinion leaders first. In some

online forums, the attributes of the members, such as the number of followers, are shown. The

public can judge whether an individual is the opinion leaders or not by their number of followers

or others directly. However, many online communities do not have such function. Although for

the insiders, it may be easy to identify the opinion leaders whom they like, for outsiders, such

as marketers, it will be difficult for them to judge the influential individuals, let along utilizing

them for marketing strategies.

Under such situation, how to identify opinion leaders in the online communities of

consumption in which they cannot be identified directly becomes a question.

Prior researches have shown that researchers have had many empirical studies on this issue

and that the approaches for identifying online opinion leaders can be summarized into three

categories: 1) the user attributes analysis; 2) the text mining analysis; and 3) network structure

analysis.

However, little research is about how to identify the opinion leaders in the online

communities of consumption, let alone in the online communities of consumption in which they

cannot be identified directly.





Hence, in order to fill in this research gap and provide theoretical and practical implications

for future study, this article is going to identify the opinion leaders in the online communities

of consumption in which they cannot be identified directly. The rest of this article is organized

as follows: in the second section, a literature review of the approaches for identifying online

opinion leaders was provided. After the discussion, SNA was selected as the approach and the

specific indicators were pointed out. In the third section, the selected sample and the process of

data collection were introduced. In the fourth section, the specific data analysis was explained

and the opinion leaders were identified. The fifth section provided the discussion and

conclusion and the final section outlined the limitation and future study.

2. Literature Review

2.1 Opinion Leader

The opinion leaders are first defined as the individuals who pass the information from the

mass media to the public with adding their own understandings (Lazarsfeld et al., 1944).

Rogers and Cartano (1962) deemed them as individuals “who exert an unequal amount of

influence on the decision of others ...those individuals from whom others seek advice and

information”.

Glock and Nicosia (1963) considered them as the individuals who act as both channels of

information and a source of social pressure toward choices.

Kotler (2001) defined them as the individuals affecting members in the social community

because of their special techniques, knowledge, personalities and other features.

2.2 Identification Approaches for Online Opinion Leaders

Prior researches show that for identifying online opinion leaders, the approaches can be

classified into three categories as follows.

1) User attributes analysis

Many researchers have analyzed the attributes of users, such as the number of concerns, the

number of fans, the number of texts and others, so as to identify online opinion leaders (Wang

et al., 2006; Ding & Wang, 2010; Ding et al., 2010; Xu & Chen, 2010; Zhu et al., 2011; Liu et

al., 2011; Li et al., 2013; Zhang et al., 2014).

2) Text mining analysis

Many researchers have focused on the contents of the posts and analyze the influences of

the words inside the contents so as to identify opinion leaders (Matsumura et al., 2002; Shi et

al., 2008; Agarwal et al., 2008; Song et al., 2012; Fan et al., 2013).

3) Network structure analysis

This approach is to analyze the structure of the network for identifying opinion leaders. The

related researches include two categories:





The first one is to utilize the classical network topology analysis, such as the PageRank, to

identify the opinion leaders. PageRank is a computing technology of web page rank to calculate

the number of hyperlinks among different web pages and thus to identify opinion leaders (Page

et al., 1998). The PageRank and the new algorithms based on it are used to identify the opinion

leaders in BBS (e.g. Ning & Wan, 2013), in a network (Zhou et al., 2009), in a blog (Song et

al., 2007), in microblog (Weng et al., 2010), in BBS (Yu et al., 2010), in twitter (Lin et al.,

2013), and so forth.

The second approach is to use social network analysis (SNA) to identify opinion leaders.

Social network analysis (SNA) refers to the process of investigating social structures

through the use of network and graph theories (Liu, 2004). It is to investigate social structures

by using network and graph theories (Wasserman & Faust, 1994).

Previous researches show that SNA is widely used to identify opinion leaders in BBS (Gao

& Yang, 2007), in blogs (Chen & Liu, 2015); in Microblogs (Wang & He, 2013; Zhao, 2013),

in college student groups (Luo & Xi, 2012), in social networks (Cho et al., 2012), in an online

community of knowledge (Wang & Ma, 2009; Liu et al., 2014).

2.3 The Approach Chosen for this Study—SNA

The literature review show that few research focuses on identifying opinion leaders in the

online communities of consumption, let alone in the online communities of consumption in

which they can not be identified directly. Hence, this part aims at having an empirical study on

identifying the opinion leaders in such online communities. With this research purpose, the user

attributes analysis and the text mining analysis are not suitable.

On one hand, the user attributes analysis is to utilize the basic attributes of individuals for

judging whether these people are opinion leaders are not. Because this research focuses on the

online communities of which fail to provide the functions of showing the attributes of

individuals so as to identify opinion leaders directly, the user attributes analysis is not suitable.

On the other hand, the text mining analysis is to utilize the texts in the posts or replies to

identify the opinion leaders, but neglects the relationships of different posts or the interactive

relationships among individuals. Since individuals are interacting with others in the online

communities of consumption, the text mining analysis is also not suitable.

For this research, because of two reasons, SNA is more suitable.

On one hand, in the traditional cases, the network analysis is also utilized to identifying

opinion leaders in the communities (Berelson & Steiner, 1964; Scott, 1992; Wasserman &

Faust, 1994).

On the other hand, using SNA has the unique advantages because that it can analyze the

pattern of interpersonal communication, scrutinizes relationships with the directions and

strength of the relationships (F. Li & Du, 2011; Valente, 1995).

Hence, this study chooses to use SNA.





When utilizing SNA to identify opinion leaders, the concepts of network location is crucial

(Liu,2004). The nodes taking central positions and the structural hole positions are always

deemed as the individuals who are more able to be opinion leaders. The nodes taking the central

positions refer to the individual actors in a social network who have many connections with

other members and who are in the center of the network. The structural hole position refers to

the location of a third party who can connect two individuals who otherwise will be separate.

However, only utilizing the central positions or the structural hole positions to judge whether

an individual is opinion leader or not is not enough (Van der Merwe & Van Heerden, 2009;

Xie, 2010).

Consequently, in this article, these two positions will be integrated for identifying the

opinion leaders. Namely, the indicators of opinion leadership used in this article include:

InDegree

OutDegree

Betweenness

inFarness and outFarness

EffSize

ConStraint

Here, the first to the fourth indicators are for measuring the degree to which one individual

is taking the central positions, and the fifth to sixth indicators are for measuring the degree to

which the individual is taking the structural hole positions. To be more specific, the InDegree

centrality is to measure the number of the replies received by the posters from the repliers and

the OutDegree centrality is to measure the number of the repliers sent by a specific node to

others. The Betweenness centrality is to measure the control degree of the node towards the

overall resources inside the network. The Closeness centrality is to measure how close a node

is to other actors in the network. The Effect Size is to measure the effective size of ego's

network. The Constraint is to measure the extent to which the ego is connected with others.

3. Research Design

3.1 The Sample

Considering that China is very representative, the sample of this thesis was chosen randomly

among the online communities of consumption in China.

To be more specific, on one hand, the scale of overall net citizens in China increases greatly.

According to the 39th China Internet Network Development State Report, published by China

Network Information Center(CNNIC), as of December 2016, the Internet users in China

increased to 731 million, increased by a total of 42.99 million throughout the year. The Internet

penetration rate was 53.2% and increased 2.9% When compared with the amount of it in 2015.





On the other hand, more and more individuals get used to make transaction and

communicate online. according to the 39th China Internet Network Development State Report,

as of December 2016, the Internet consumers in China increased to 467 million, which took up

63.8% of the total number of Chinese netizens.

Hence, this article randomly selected Changsha Tong, an online community of consumption

in China and this online community also meet with the requirement that the opinion leaders

cannot be identified directly. This community is divided into six parts, including the parts about

shopping, wearing, panic purchasing, parenting, foods and group buying. The shopping part is

the biggest part. The members inside this community are sharing their shopping experiences

and leaving recommendations of products or services. According to its introduction, its average

Page View per day reached 200 thousand, and IP per day reached 30 thousand in the August,

2013. Until August, 2013, the registered members reached 530 thousand, and they were mainly

aging from 23 years old to 35 years old.

3.2 Data Collection

This article selected the members in the shopping part of the Changsha Tong as the research

target. Since it is different to get the details of all the members inside the whole website, this

article chose to use the snowball sampling to collect the data of the members and selected the

posts in a week as the data sources.

In one week, from 26th, Dec, 2016 to 1st, Jan, 2017, this shopping part had 1193 pieces of

new posts. The posts related to shopping was selected and others were deleted, such as the posts

only for chatting. Finally, 408 posts were selected and 1024 members were involved.

Taking the large number of members of Changsha Tong into account, the number of 408

indicates that most members were silent. Without leaving messages, the silent members are

considered to be unlikely to affect others. Hence, lacking the data of these members will not

affect the results of this article.

4. Data Analysis

4.1 Analysis of Degree Centrality

By using Ucinet 6, the results of Degree centrality are shown in the Table 1.

Table 1: Analysis of Degree Centrality

Number ID 1 2 3 4

OutDegree InDegree NrmOutDeg NrmInDeg

150 蔡大婶 96 81 0.722 0.609

152 沐小言 96 79 0.722 0.594

93 大小眼 79 0 0.594 0

174 咦～嘿嘿 79 0 0.594 0

175 泡泡不是彩色的 72 71 0.541 0.534

128 跟屁虫 67 161 0.504 1.211





120 微微筱 67 42 0.504 0.316

29 Amy1234 64 19 0.481 0.143

18 铃铛小亚 63 75 0.474 0.564

95 yachne 60 0 0.451 0

Notes. (Limited by the space, only the information of first 10 members is shown)

In the table, the analysis of every member is shown. The first column in the table indicates

the original number of the members and the second column indicates the ID of the members.

The third column is the Outdegree of the members and the four column is the InDegree of them.

The fifth column is the NrmOutDeg, which represents the standard value of the OutDegree and

the sixth column is the NrmInDeg, which represents the standard value of the InDegree. Here,

the higher the InDegree is, the more prestigious the member is and the more he or she is in the

core position.

Because the minimum of OutDegree and of InDegree were all equal to 0, it indicates that

the isolated nodes existed. The standard deviation of OutDegree was 10.432, much smaller than

the standard deviation of InDegree. It indicates that the values of OutDegree was not so

different from the mean of it and the values of InDegree was comparatively different from the

mean of InDegree. It also indicates that the numbers of the posts were even, but the numbers of

received replies were uneven. Namely, there were some posts which received more pieces of

comments than other posts.

4.2 Clustering Analysis of the Indegree and of the Outdegree

By using SPSS 23 to have a clustering analysis of the InDegree and OutDegree of the

members, the results are shown in the Table 2 and Table 3.

Table 2: Part 1 of Clustering analysis of the InDegree and of the OutDegree

Clustering Analysis

Clustering

1 2 3 4 5

InDegree 27.93 70.24 2.19 291.00 164.33

OutDegree 66.80 28.72 3.73 48.00 45.33

Table 3: Part 2 of Clustering analysis of the OutDegree and of the InDegree

Number of nodes in every cluster

clustering 1 15.000

2 29.000

3 976.000

4 1.000

5 3.000

effective 1024.000

missing .000





Through the clustering analysis, the members were divided into 5 categories.

The first one included 15 members. They had high values of OutDegree, but their InDegree

were not comparatively high. The InDegree centrality represents the number of the replies

received by the posters from the repliers, and the OutDegree centrality represents the number

of the repliers sent by a specific node to others. Hence, these 15 members always replied others,

but they did not reply so much replies.

The second one included 29 members. They had high values of InDegree and of OutDegree.

Hence, these 29 members always replied other and got many replies. Their relationships with

some other members were comparatively strong.

The third one included 976 members, which took up 95% of the total members. These

members had low values of InDegree and of OutDegree. Hence, they did not reply many posts

and they receive a few replies. These members were lacking of the interaction with others.

The fourth one included only 1 member. Compared to all other members, this member had

the highest values of InDegree and of OutDegree. Hence, this member replied a lot and received

the most replies.

The fifth one included 3 members. They had relatively high values of InDegree and their

values of OutDegree were also high. Hence, these 3 members replied many members and

received the comparatively high number of replies.

All in all, compared to other members, the members in the first, second, fourth and fifth

categories interacted more with others and were comparatively active. These 34 members may

be the opinion leaders. The number of them were: “380”, “23”, “128”, “20”, “56”, “2”, “113”,

“225”, “267”, “77”, “188”, “41”, “150”, “152”, “18”, “92”, “13”, “175”, “3”, “162”, “415”,

“420”, “400”, “114”, “619”, “52”, “270”, “418”, “106”, “88”, “121”, “173”, “189”, and “394”.

Their ID were : “苏大大大大大”, “康康砣”, “跟屁虫”, “临去秋波”, “败家的甘小妞妞”, “倩

倩砣”, “小小樱桃”, “miss 柚子”, “白将军”, “潇湘猫猫”, “芳芳”, “莫小喵”, “蔡大婶”, “沐

小言”, “铃铛小亚”, “雅笛”, “yyyyyyyyy”, “泡泡不是彩色的”, “dengya_wen”, “夏沫冬至”,

“懒 xiao 白”, “五五呀”, “木槿花微眠”, “安南妈”, “A 灰太狼 a”, “Soga”, “蜜蜂 fffff”,

“fish727”, “Sunshy7”, “十八大姨妈”, “俊妈咪小 P”, “晶晶 dsuger”, “5.27 向日葵”, and “翁

羽西”.

4.3 Analysis of Betweenness Centrality

By using Ucinet 6, the results of Betweenness centrality are shown in the Table 4.

Table 4: Analysis of Betweenness

Number ID 1 2

Betweenness nBetweenness

380 苏大大大大大 154350.547 14.763

23 康康砣 121755.336 11.646

128 跟屁虫 87438.781 8.363





20 临去秋波 85143.242 8.144

188 芳芳 68944.75 6.594

164 dark_darky 57067.199 5.458

174 咦～嘿嘿 51500.48 4.926

77 潇湘猫猫 51276.117 4.904

41 莫小喵 49169.605 4.703

225 miss 柚子 47165.516 4.511




The third column is the Betweenness of the members and the four column is the nBetweenness

of them. From the table, it can be seen that the top one was the member called “苏大大大大

大”. The Betweenness of this member was 154350.547 and it means that this member was on

the shortcut to many nodes inside the network. Namely, this member usually served as an

intermediary member and many information was flowed to others through this member. Hence,

this member had a very important position inside the network.

Meanwhile, there were many members who have the Betweenness being 0 (the data was

not shown in this table). It is between that these members were in the position of hanging out,

namely they just had relationships with one member inside the communities. They were not in

the shortcut of any two members and thus their values of Betweenness was 0.

4.4 Analysis of Closeness Centrality

By using Ucinet 6, the results of Closeness centrality are shown in the Table 5.

Table 5: Analysis of Closeness Centrality

Number ID 1 2 3 4

inFarness outFarness inCloseness outCloseness

821 熙、小贝 5252 1047552 19.478 0.098

380 苏大大大大大 5255 2195 19.467 46.606

128 跟屁虫 5285 2260 19.357 45.265

150 蔡大婶 5298 2251 19.309 45.446

164 dark_darky 5326 2270 19.208 45.066

152 沐小言 5349 2312 19.125 44.247

174 咦～嘿嘿 5365 2310 19.068 44.286

161 飞飞 1815 5405 2379 18.927 43.001

29 Amy1234 5406 2346 18.923 43.606

2 倩倩砣 5429 2378 18.843 43.019




The third column is the inFarness of the members and the four column is the outFarness of

them. The inFarness represents the sum of the distance from this node to other information

receivers and the outFarness represents the sum of the distance from the information senders to

this node. From the table, the member called “苏大大大大大” had the shortest distance to





other members. It means that this member might be in contact with most of the members of the

network or be able to contact the other members of the network at a very short distance. It also

can be seen that this member had the minimal control when they want to get information from

other members.

4.5 Analysis of Structure Holes

By using Ucinet 6, the results of structure holes are shown in the Table 6.

Table 6: Analysis of Structure Holes

No. ID Structural Hole Measures

1 2 3 4 5 6 7 8 9 10

Degree EffSize Efficienc Constrain Hierarchy EgoBet

Ln(Co

nstr

Indirec

ts

Densit

y Numholes

1 满天都是

馨馨 17.000 14.401 0.847 0.200 0.165 73.833 -1.611 0.605 0.195 219.000

2 倩倩砣 88.000 82.611 0.939 0.054 0.177 2451.707 -2.927 0.752 0.072 7102.000

3 dengya_w

en 64.000 58.932 0.921 0.079 0.190 1105.298 -2.534 0.793 0.097 3639.000

4 老傅 9.000 8.333 0.926 0.249 0.230 0.000 -1.389 0.311 0.083 66.000

5 teroyo 4.000 3.667 0.917 0.392 0.006 0.000 -0.936 0.250 0.167 10.000

6 lovejamt 15.000 13.022 0.868 0.202 0.123 0.000 -1.601 0.589 0.205 167.000

7 xixihaha86 5.000 3.400 0.680 0.700 0.441 0.000 -0.356 0.511 0.300 14.000

8 细鱼嫩子 8.000 7.692 0.962 0.207 0.182 23.000 -1.575 0.154 0.036 54.000

9 chy901 1.000 1.000 1.000 1.000 1.000 0.000 0.000 0.000 0.000

10 zcx 1.000 1.000 1.000 1.000 1.000 0.000 0.000 0.000 0.000


From the Table 6, the analysis of every members is shown. The first column in the table

indicates the original number of the members and the second column indicates the ID of the

members. From the third column, there are some structural hole measures. The most important

and the second important values are the EffSize and Constrain. The EffSize represents the

nonredundant contacts of a node. The higher the EffSize is, the more nonredundant contacts

does the node have. Meanwhile, the ConStraint measures the extent to which the node’s

network limits his or her time and energy. The lower the Constraint is, the opener the network

will be and the more possible to have structural holes.

4.6 Analysis of the indicators’ data and identification

From the section 4.2, the result of the clustering analysis of the OutDegree and of the

InDegree has shown that 34 members might be the opinion leaders. Hence, the calculated data

of these 34 members were listed in the Table 7 and the possibility of their being opinion leaders

were discussed.

Table 7: Analysis of the indicators’ data

No. ID In

Degree

Out

Degree

EffSize Constrain Betweenness inFarness outFarness

380 苏大大大

大大

291 48 161.921 0.032 154350.547 5255 2195

23 康康砣 189 33 117.542 0.037 121755.336 5430 2375

128 跟屁虫 161 67 110.211 0.045 87438.781 5285 2260

20 临去秋波 143 36 125.176 0.032 85143.242 5538 2442

56 败家的甘

小妞妞

128 22 67.745 0.077 38835.934 5614 2556





2 倩倩砣 121 57 82.611 0.054 45835.426 5429 2378

113 小小樱桃 114 33 73.079 0.057 5541 2488

225 miss 柚子 109 23 73.295 0.064 47165.516 5566 2512

267 白将军 102 36 71.422 0.067 39184.133 5511 2425

77 潇湘猫猫 99 34 71.635 0.056 51276.117 5572 2508

188 芳芳 90 34 75.384 0.044 68944.75 5507 2466

41 莫小喵 87 15 66.755 0.05 49169.605

150 蔡大婶 81 96 74.514 0.057 40815.316 5298 2251

152 沐小言 79 96 64.986 0.069 34886.75 5349 2312

18 铃铛小亚 75 63 54.672 0.077 25664.84 5462 2411

92 雅笛 73 38 55.72 0.075 30097.926

13 yyyyyyyy

y

72 50 42.767 0.095 15683.476 5483 2418

175 泡泡不是

彩色的

71 72 58.281 0.071 20294.611 5462 2402

3 dengya_w

en

71 48 58.932 0.079 28336.488 5499 2441

162 夏沫冬至 65 58 45.142 0.083 15418.374 5435 2380

415 懒 xiao 白 64 12 32.767 0.129

420 五五呀 62 24 46.424 0.071 16388.652

400 木槿花微

眠

60 17 56.157 0.067 40839.59 5597 2530

114 安南妈 59 24 48.051 0.093 22950.369

619 A 灰太狼a

59 6 33.579 0.084

52 soga 58 10 42.816 0.059 21221.381

270 蜜蜂 fffff 55 44 42.159 0.089 5581 2533

418 fish727 53 7 25.704 0.122 17923.656

106 Sunshy7 51 19 39.029 0.374

88 十八大姨

妈

49 19 28.004 0.137

121 俊妈咪小P

47 10 29.88 0.143

173 晶晶dsuger

46 17 29.632 0.101

189 5.27 向日

葵

44 34 37.015 0.116 16618.361

394 翁羽西 44 13 35.524 0.079 16905.506

Notes. (*Table 4 lists the information of Betweenness of first 40 members. Table 5 lists the information

of both inFarness and outFarness of first 40 members. If the 34 members’ ID were not shown in the Table

4, their Betweenness are not reported in the table above. Similarly, if the 34 members’ ID were not shown

in the Table 5, their inFarness and outFarness are not reported in the table above.)

From the table 7, the opinion leaders inside this community could be identified.

Nodes with numbers of “380”, “23”, “128”, “20”, “56”, “2”, “225”, “267”, “77”, “188”,

“150”, “152”, “18”, “13”, “175”, “3”, “162” and “400” could considered as the opinion

leaders.

Their were “苏大大大大大”, “康康砣”, “跟屁虫”, “临去秋波”, “败家的甘

小妞妞”, “倩倩砣”, “miss 柚子”, “白将军”, “潇湘猫猫”, “芳芳”, “蔡大婶

”, “沐小言”, “铃铛小亚”, “yyyyyyyyy”, “泡泡不是彩色的”, “dengya_wen”,

“夏沫冬至” and “木槿花微眠”.





Because they had the comparatively highest values of InDegree, OutDegree, EffSize,

Betweenness and lowest values of Constrain, inFarness and outFarness, they could be

considered as the opinion leaders in this online community.

5. Discussion and Conclusion

Opinion leaders is considered to affect the flow of eWOM information and the purchase

choice of consumer through eWOM communication greatly. Hence, how to identify the online

opinion leaders has been one of the hottest questions for researchers. For this research, the

opinion leaders in the online communities of consumption in which they cannot be identified

directly were focused.

This article begins with a literature review of the approaches of identifying the online

opinion leaders and pointed out that SNA is suitable for identifying the online opinion leaders

in the social network. Then an empirical study by utilizing SNA to identifying the randomly

selected online community of consumption in China is held. The results show that 18 opinion

leaders in this communities are identified.

Theoretically, this empirical study provides one of the first empirical studies towards the

application of SNA on identifying the opinion leaders inside the online communities of

consumption in which they cannot be identified directly and fill in the research gap that little

researches focus on identify the online opinion leaders in such online communities.

Practically, for the organizers of online communities of consumption or marketers, they can

utilize SNA to identify the opinion leaders in those communities which fail to distinguish

opinion leaders from other members at first. Then they can make full use of these opinion

leaders for the marketing strategies.

6. Limitation and Future Study

Although the author collected and analyzed the relevant researches and had the empirical

analyses, the limited time, limited resources, limited research abilities and others lead to the

limitations of this thesis. It is these limitations that provides the directions for future researches.

The limitations include: 1) The randomly selected online community of consumption in this

article may fail to represent all other online communities of consumption. 2) The sample of

China may not enough to explain all the situations in other countries. 3) The time period of data

was only one week and the opinion leadership may change in the long time. 4) The snowball

sampling is useful for collecting data in a short time, but it will be difficult to collect data in a

long-time period. 4) The indicators for identifying online opinion leaders including the

indicators of central positions and of structure holes, are actually subjective, out of the author’

choice.

Consequently, future research can improve the samples by investigating more online

communities, choosing other counties and collecting the data in the long-time period.





Meanwhile, future research can choose some efficient approaches to collect data, such as

designing some algorithms. Furthermore, future research can improve the research method by

not only limited to calculating the indicators of central positions and of the structure hole

position.

References

Agarwal, N., Liu, H., Tang, L., & Yu, P. S. 2008, Identifying the influential bloggers in a community.

Proceedings of the 2008 international conference on web search and data mining.

Berelson, B., & Steiner, G. A. 1964, Human behavior: An inventory of scientific findings.

Chen, Y., & Liu, X. 2015, Recognizing opinion leaders based on Social Network Analysis. Journal of

information Science, 33, 13-20, (In Chinese).

Cho, Y., Hwang, J., & Lee, D. 2012, Identification of effective opinion leaders in the diffusion of

technological innovation: A social network approach. Technological Forecasting and Social Change,

79(1), 97-106.

Ding, H., & Wang, Y. 2010, An analysis of characteristics of opinion leaders in SNS -- a case study of

Douban website. Journal of News and Communication Research (3), 82-91, (In Chinese).

Ding, X., Hu, Y., Zhao, W., Wu, R., Hu, C., & Yang, Y. 2010, A study on the characteristics of online

opinion leaders. Journal of Sichuan University (2), 145-149.

Fan, X., Zhao, J., Fang, B., & Li, Y. 2013, Influencing diffusion probability model and its application

for opinion leader discovery. Journal of Computer Science, 36(2), 360-367.

Gao, J., & Yang, J. 2007, Analysis on the opinion leaders in the online forums. Journal of university of

electronic science and technology of China, 36(6), 1249-1252, (In Chinese).

Glock, C. Y., & Nicosia, F. M. 1963, Sociology and the Study of Consumers. Journal of Advertising

Research, 3(3), 21-27.

Goldsmith, R. E., Flynn, L. R., & Goldsmith, E. B. 2003, Innovative consumers and market mavens.

Journal of Marketing theory and practice, 11(4), 54-65.

Hagel, J., & Armstrong, A. G. 1997, Net gain. Harvard Business School Press, Boston.

Kotler, P. 2001, Marketing Management, 10: Pearson Education Canada.

Kozinets, R. V., De Valck, K., Wojnicki, A. C., & Wilner, S. J. 2010, Networked narratives:

Understanding word-of-mouth marketing in online communities. Journal of marketing, 74(2), 71-89.

Lazarsfeld, P. F., Berelson, B., & Gaudet, H. 1944, The people\'s choice; how the voter makes up his

mind in a presidential campaign.

Li, F., & Du, T. C. 2011, Who is talking? An ontology-based opinion leader identification framework

for word-of-mouth marketing in online social blogs. Decision Support Systems, 51(1), 190-197.

Li, C.-Y. 2013, Persuasive messages on information system acceptance: A theoretical extension of

elaboration likelihood model and social influence theory. Computers in Human Behavior, 29(1), 264-

275, (In Chinese).

Lin, Y., Li, H., Liu, X., & Fan, S. 2013, Hot topic propagation model and opinion leader identifying

model in microblog network. Paper presented at the Abstract and Applied Analysis.

Liu, Z., & Liu, L. 2011, Identification and analysis of opinion leaders in Micro-blog. Journal of Systems

Engineering, 29(6), 8-16, (In Chinese).

Luo, X., & Xi, l. 2012, Research on opinion Leaders of electronic word-of-mouth based on Social

Network Analysis. Journal of E-commerce and Information Management, 24(1), 75-81, (In Chinese).

Matsumura, N., Ohsawa, Y., & Ishizuka, M. 2002, Mining and characterizing opinion leaders from

threaded online discussions. Proceedings of the 6th International Conference on Knowledge-Based

Intelligent Engineering Systems & Allied Technologies.





Ning, L., & Wan, Z. 2013, Identifying opinion leaders Based on Group Review. Information magazine,

32(8), 204-206, (In Chinese).

Page, L., Brin, S., Motwani, R., & Winograd, T. 1998, The PageRank citation ranking: Bringing order

to the web.

Rogers, E. M., & Cartano, D. G. 1962, Methods of measuring opinion leadership. Public Opinion

Quarterly, 435-441.

Rogers, E. M. 2003, Diffusion of innovations. Free Press. New York, 551.

Shi, M., Fang, Y., Zeng, X., & Wang, C. 2008, Analysis of IDM model and its reformative arithmetic.

Journal of Chengdu University of Information Technology, 23(1), 69-72, (In Chinese).

Song, K., Wang, D., Feng, S., Wang, D., & Yu, G.(2012, Detecting Positive Opinion Leader Group from

Forum.

Song, X., Chi, Y., Hino, K., & Tseng, B. 2007, Identifying opinion leaders in the blogosphere.

Proceedings of the sixteenth ACM conference on Conference on information and knowledge

management.

Scott, J. 1992, Network analysis: A handbook: Sage Publications.

Stern, B. B., & Gould, S. J. 1988, The consumer as financial opinion leader. Journal of Retail Banking,

10(2), 43-52.

Summers, T., King, E., Martin, D., & Jackson, R. 1976, Biological control of Diatraea saccharalis in

Florida by periodic releases of Lixophaga diatraeae. BioControl, 21(4), 359-366.

Valente, T. W. 1995, Network models of the diffusion of innovations.

Valente, T. W., & Pumpuang, P. 2007, Identifying opinion leaders to promote behavior change. Health

Education & Behavior, 34(6), 881-896.

Van der Merwe, R., & Van Heerden, G. 2009, Finding and utilizing opinion leaders: Social networks

and the power of relationships. South African Journal of Business Management, 40(3), 65-76.

Wang, D., & He, Y. 2013, Analyzing structure of opinion leaders based on Social Network Analysis.

Journal of Statistics and information forum, 28(10), 84-89.

Wang, L., & Ma, R. 2009, The function of opinion leaders in the online communities of knowledge.

China academic journal of electronic publishing house, 1, 54-58, (In Chinese).

Wang, X., Li, X., & Chen, G. 2006, Complex network theories and the applications, Qinghua University

Publishing, (In Chinese).

Wasserman, S., & Faust, K. 1994, Social network analysis: Methods and applications (Vol. 8):

Cambridge university press.

Weng, J., Lim, E.-P., Jiang, J., & He, Q. 2010, Twitterrank: finding topic-sensitive influential twitterers.

Proceedings of the third ACM international conference on Web search and data mining.

Xie, Y. 2010, Research on the influences from network positions inside Blogs. Master thesis of Yangzhou

University, (In Chinese).

Xu, K., & Chen, X. 2010, Research on the model for the influences of "opinion leaders" in BBS.

Journalism Quarterly, 4, 87-93, (In Chinese).

Yu, X., Wei, X., & Lin, X. 2010, Algorithms of BBS Opinion Leader Mining Based on Sentiment

Analysis. WISM, 10, 360-369.

Zhang, W., Li, X., He, H., & Wang, X. 2014, Identifying network public opinion leaders based on markov

logic networks. The scientific world journal.

Zhao, H. 2013, Research on the identification of opinion leaders in Micro-blog based on Social Network

Analysis. Journal of E-commerce and Information Management (9), 63-64, (In Chinese).

Zhou, H., & Zeng, D. 2009, Finding leaders from opinion networks. Proceedings of the Intelligence and

Security Informatics, 2009. ISI'09.

Zhu, S., Zheng, X., & Chen, D. 2011, Research on the algorithm of automatic discovery of opinion

leaders in forums. Journal of System Engineering Theory and Practice, 31(S2), 7-12.

an empirical study on identifying opinion leaders in...

Documents