Examining the Determinants of Who is Hyperlinked to Whom: A Survey of Webmasters in Korea by Han Woo Park
This paper assesses the determinants of who is hyperlinked to whom on the Web. It details the results of a conducted survey with 64 Korean Webmasters to examine the reasons Web sites form hyperlink networks with other sites. The results indicated that Webmasters consider the credibility of hyperlinked Web sites to be above average when deciding to hyperlink. Particularly, the item "usefulness" had the highest score, indicating that Web sites are more likely to hyperlink to Web sites possessing practical content, information, or services. Additionally, the qualitative analysis reveals that hyperlinking motivations and advantages are largely grouped into two dimensions: navigational functionality and for business purposes.Contents
Hyperlink Research: Overview and Methodological Aspects
Determinants of Hyperlinking: Web Site Credibility
Research Question
Methods
Results
Discussion
Limitations and Future Research
Hyperlink Research: Overview and Methodological Aspects
The World Wide Web is a systematic body of Web sites that contain information. A hyperlink between two Web sites keeps the two sites together. Hyperlinking is a technological capability that enables one specific Web site to link directly to another. There has been a surprising growth of hyperlink analysis studies in the Internet community (Park, in press). Hyperlink analysis examines Web interactivity, international and interpersonal communications, and e-commerce, among other topics.
While any individual or institution has complete freedom in choosing the direction of hyperlinks on their Web sites (or Web pages), Albert et al. (1999) demonstrated that the Web possesses a "flocking" nature; you can get from one Web page to another by clicking on hyperlinks an average of 19 times, if you select two Web pages at random. While Albert et al. (1999) did not clarify the nature of directionality on the Web, Broder et al. (2000) calculated an average distance of the Web considering the direction. They examined about 200 million pages and 1.5 billion links. Their study revealed that more than 90 percent of the sample Web pages form a single connected part if links are treated as a unidirectional link. About 56 million pages are connected to one another along directed links. The probability that there is a hyperlink between two randomly chosen Web sites is only 24 percent. When there is a link, in-links and out-links are 16.12 and 16.18 respectively. The figure can be much smaller, 6.83, if an undirected path exists. We should note that it is not a geometric distance but the pattern of connections, or rather, the topology (Hayes, 2000).
Halavais (2000) examined the role of geographic borders in cyberspace using the hyperlink patterns of Web sites. Specifically, he took a sample of 4,000 Web sites and analyzed their external hyperlinks, determining the total percentage of hyperlinks from the sites to various countries. In a study on the structure of global commerce on the Web, Brunn and Dodge (2001) used a similar method to analyze the inter-domain hyperlinks among 174 geographic TLDs (top-level domains, such as .ca for Canada). They treated Web sites' incoming and outgoing links separately and developed a domain-by-domain matrix of inter-hyperlinks upon which they conducted descriptive statistics and cross-tabulation analysis by country and region.
Barnett et al.'s (2001) study differed in that it used secondary data published by the Organization for Economic Cooperation and Development (OECD, 1998). The data included the number of hyperlinks embedded in Web sites between all TLDs among OECD member countries. They examined international information flow among OECD countries in terms of hyperlinkage traffic. Hyperlink analysis of this sort focused on the importance of geography to the creation of a hyperlink network.
In the context of interpersonal communication, Park et al. (2000) analyzed hyperlinks among the Web sites of Korea's National Assemblymen. They developed a site-by-site matrix of hyperlinks upon which they conducted cluster analysis. They found that the structure of their hyperlink network is significantly related to party membership.
Adamic and Adar (2001) focused on university students' (Stanford University and the Massachusetts Institute of Technology) home pages and described hyperlink connections among them. They found that some students had more than 30 incoming and outgoing hyperlinks while some of their schoolmates did not have any links at all. In order to locate the connector that played a key role in linking homepages in the university, they measured the average of the shortest path between any two homepages (9.2 for the Stanford network and 6.4 for the MIT). They suggested that these results might reflect the existence of a small world network online as well as offline (Milgram, 1967; Watts and Strogatz, 1998).
The Internet has changed the way in which scholars communicate and collaborate (Borgman and Furner, 2001). It allows for the synchronization and multiple linking among scholars, journals, and scientific organizations through hyperlinks embedded in Web sites. Hyperlinks function as a new channel for formal and informal scholarly communication. Based on a focus interview with 15 scholars who included external hyperlinks in their academic papers, Kim (2000) explored influences on hyperlinking behavior. Overall, the hyperlinks in scholarly electronic articles are motivated by a variety of reasons. However, he was able to develop a typology of scholars' motivations for hyperlinking. The motivations for this new citation, called a "sitation", are grouped into three reasons (scholarly, social, and technological). Besides technological reasons, other motivations were akin to those found in traditional scholarly environments.
Palmer et al. (2000) employed hyperlink analysis to examine e-commerce. They used the number of inward hyperlinks to a Web site as an indicator of trust placed in Internet firms. They obtained data from Alexa.com. As a Web information company, Alexa provides a variety of statistics about individual Web sites. The results revealed that the number of incoming links was strongly related to the use and prominence of TTPs (Trusted Third Parties); privacy statements that may be also used as a trust indicator.
Kreb's (2000) study of Amazon.com indirectly revealed the role of hyperlinks in relation to homophilous attribute among online consumers. Homophily is defined as "the degree to which pairs of individuals who interact are similar in certain attributes, like beliefs, values, education, and social status" [ 1]. In other words, the more homophilous people are in social and attitudinal characteristics the more frequently communication occurs. This is commonly known as the adage "birds of a feather flock together." Amazon.com provides its customers with information about books similar in nature purchased by other customers. It contains a hyperlink so that prospective customers can take a look at these hyperlinked books directly. He argues that a consumer could be persuaded to buy a certain book knowing that other customers also purchased the same book. Park et al. (2002) regarded that the number of incoming hyperlinks to a Web site could be used as an indicator of perceived credibility; they analyzed the affiliation network among Korea's most frequently visited 152 Web sites. They created a Web site-by-Web site relations matrix based on the existence of hyperlinks in a Web page affiliation program. Web sites that did not play a significant role in the hyperlink network (e.g., isolate) were excluded. They found that the structure of the affiliation network was influenced by the financial Web sites with which others were affiliated.
Determinants of Hyperlinking: Web Site Credibility
As mentioned earlier, hyperlink research has examined social and communication structures among different groups - individuals, organizations or nation-states - on the Web. However, there has been insufficient research examining what influences hyperlinking among Web sites. In other words, little research has been done to investigate the determinants of hyperlink relations among Web sites.
What motivates a Web site to hyperlink to another Web site? Given that many Web sites are still in an early stage of development, a partner Web site's credibility may be a critical element in deciding with whom to hyperlink. The main purpose of a Web site is to accumulate and sustain traffic to the site. Web behavior - visiting a specific Web site to seek information or purchase a product - may be thought of as a form of persuasion. It entails a change in one's attitude or behavior over time. In the field of communication, a communicator's credibility - or in this case, a Web site's credibility - is regarded as one of the most influential factors in the persuasion process (Berlo et al., 1969; Hovland et al., 1953; Hovland and Weiss, 1951; McCroskey and Teven, 1999; Sundar and Nass, 2001). Recently, communication researchers have turned their attention to Web site (or Internet) credibility (Flanagin and Metzger, 2000; Johnson and Kaye, 1998; Kiousis, 1999; Park et al., 2002; Schweiger, 2000; Tseng and Fogg, 1999).
Each dimension of Web site credibility may be defined in the following statements. Trustworthiness refers to the truthfulness of a Web site's contents and the site's reputation. Expertise is the Web site's apparent competence in how complete, useful, and timely the services of the Web site are when compared to others. Safety is defined as how secure or reliable the Web site's technical systems are for online payment and personal information.
Once users perceive that a Web site lacks credibility, they are likely to stop visiting it or perform financial transactions (Gefen, 2000; Park et al., 2002; Tseng and Fogg, 1999). Thus, one can infer that Web sites would prefer to select partners with high credibility (Evans and Wurster, 1999; Hagel and Armstrong, 1997; Shapiro and Varian, 1999). Terveen and Hill (1998) examined the number of hyperlinks between Web sites as an indicator of the quality of sites; they found that hyperlink connectivity had a significant relationship to quality judgments of sites. The indegree connectivity of a site was positively correlated with these judgments. Indegree refers to the number of links a Web site receives from other sites, while outdegree is the number of links originating at a site (Park et al., 2002). In other words, indegree connectivity is defined as the number of Web sites that are linked to a given site. Through a hyperlink, an individual Web site could influence another Web site's trust, prestige, authority or credibility (Kleinberg, 1999a; 1999b). Measuring the credibility of a site - based on hyperlink analysis - is becoming a more reliable method than using the number of hits or visitors a site receives (Henzinger, 2001; Ingwersen, 1998).
Research Question
The decision to form hyperlinks with certain Web sites is in part subject to a Webmaster's perception of the credibility of those sites. Webmasters represent Web sites (Jackson, 1997); they communicate with each other on behalf of their sites. For example, in networks where linkages are influenced by commercial reasons, Webmasters ask permission before linking their sites to another site. Although the Internet strategy of including the inherent characteristics of an organization in its Web site is not always determined by the Webmaster (Gretzel, 2001), the planning of where, how, and when to hyperlink is often controlled by the Webmaster. Therefore, perceptions of credibility are expected to be positively associated with the intentions of hyperlink exchanges. In this context, the following research question is raised:
In creating hyperlinks, how important is credibility in terms of (1) trustworthiness; (2) expertise; and, (3) safety?When considering the dominant role of Webmasters in creating hyperlinks with certain Web sites, to our knowledge, no study has clearly documented the determinants of who is hyperlinked to whom from the perspective of Webmasters. What makes Web sites form hyperlink networks with others on the Internet? What elements influence the association among Web sites via hyperlinks? This paper examines these issues. Additionally, regarding Web site credibility, past research has mainly focused on perceptions of credibility by Internet users, not content creators and producers.
Methods
Research Setting
Surveys of Internet users and Webmasters were conducted in Korea. The number of Koreans using the Internet has increased rapidly: 0.14 million in 1995, 1.6 million in 1997, 10 million in 1999, 19.04 million in 2000, and 22.23 million in September 2001 [ 2]. Korea ranks No. 1 in the Asia Pacific Region in the penetration of broadband access in 2001. Ninety-five percent of home Internet users, or 15.8 million people, have high-speed connections such as cable modem (Korea Herald, 2001). According to Nielsen/NetRatings (2001), 23 Korean Web sites are included in the list of the world's 100 biggest locally accessed Web properties. This suggests that Korean Web sites represent an adequate sample for Internet research. But the extent of using Korean Web sites to form generalizations about the rest of the world may be limited. By the same token, the degree that we can draw conclusions about other Web sites based on hyperlinking information from 64 Webmasters may not be large. The issue of generalization will be examined later in this paper.
The Data: Webmasters
In order to examine the credibility perceptions of Webmasters, a Web-based survey of 64 Webmasters was conducted from April to June 2001. The selection of Korean Webmasters was obtained by using non-random, or rather, selective sampling. Specific procedures employed in the study were as follows. First, 150 Web sites were purposely chosen from the search engine, Yahoo! Korea ( kr.yahoo.com), based on the type of Web site and its purpose. Next, a multiple-mode survey technique was used to increase response rate (Sheehan, 2001). We sent Webmasters a traditional invitation along with a questionnaire and informed them they were able to participate in the survey by either visiting a Web site ( http://www.acsu.buffalo.edu/~hanpark/webmaster_survey) or by sending a fax or e-mail. For Webmasters who did not respond, a follow-up contact, via e-mail or telephone, was completed. As resulted, 64 Webmasters participated in the survey.
Instrument Construction
The survey of Webmasters was based on eight topics: (a) a reason to choose a hyperlink with a certain Web site; (b) the advantage of hyperlinks; (c) trustworthiness of Web sites hyperlinked; (d) expertness of Web sites hyperlinked; (e) safety of Web sites hyperlinked; (f) judgment of their Web site's credibility increase when linking with other sites perceived highly credible; and, (g) affiliation information. The first, second, and last queries were asked as open-ended questions. The three dimensions of other Web sites' credibility were measured using a fractionation scale (Barnett et al., 1981; Woelfel and Fink, 1980). For example, trustworthiness was defined as the truthfulness of services such as contents. Subjects were asked to score how much they judged a Web site's services in terms of truthfulness. On this scale, 50 represented an "average" rating of truthfulness and zero represented none. Compared to a five-point Likert type scale commonly used in social science, a fractionation scale has sufficient variances to differentiate among Web sites (Barnett et al., 1981; Woelfel and Fink, 1980).
Survey Questionnaire
The survey questionnaire was originally written in English, making it possible to use the same survey questionnaire in countries other than Korea. In order to conduct a survey of Webmasters in Korea, standard back-translation procedures were employed to translate the questionnaire from English to Korean. Only Korean questionnaires were given to Korean Webmasters. Specifically, the back-translation procedure used in this research operates as follows. First, the survey questionnaire is written in English. Next, it is translated into the second language, in this case, Korean, by a second person and then, a third person translates the questionnaire back into English. Lastly, the English and Korean versions are compared by a bilingual who modifies the Korean version based on the comparison. The back-translation procedure has shown the best results in past research, by eliminating the variance between original language and translation (Brislin, 1972; Brislin and Sinaiko, 1973; Ervin and Bowers, 1952; Ervin and Osgood, 1954).
Analysis Procedure
Information collected from the survey was coded and entered into a statistical program (SPSS). Descriptive statistics were analyzed, including mean and standard deviations for each item. Factor analysis was used to explore the three underlying dimensions (trustworthiness, expertness, and safety) composing Web site credibility. Due to a small sample size, only exploratory factor analysis was done. Confirmatory factor analysis was not an appropriate method. Open-ended questions (a reason to choose a hyperlink with certain Web sites and the advantage of hyperlinks) were translated into English. They were content analyzed using CATPAC.
CATPAC (Woelfel, 1998) is a neural network computer program optimized for analyzing text. Unlike traditional (non-computerized) content analysis (Danowski, 1993; Krippendorff, 1989), CATPAC does not require prior categories. Instead, it allows categories to emerge from the data, resulting in this case as open-ended questions. In other words, CATPAC enables large bodies of text to form meaningful conceptual groupings.
The procedure of analyzing the text operates in the following manner (Woelfel, 1998). CATPAC assigns a neuron to each major word in the text. It then runs a scanning window through the text. The neuron representing a word becomes active as long as the word remains in the window. Up to n words can be in the window at once, where n is a parameter set by the user. The default value is 7; that is about as many words a human reader can look at in one glance. For example, if n is set to 7, the window will first scan words 1 through 7, then 2 through 8 and so on until the entire text is scanned. Whenever two or more words simultaneously appear in the window, the connection between them is strengthened. Mathematically, CATPAC then creates a word by word matrix, S, with each cell, sij, containing the likelihood that the occurrence of one word will trigger the occurrence of another word. Each cell, sij, indicates the degree of association between pairs of words, i and j. In other words, the cells represent the similarity of words' co-occurring in consecutive windows. Based on these procedures, CATPAC is able to conduct the following analyses.
Word Frequency Analysis
CATPAC identifies the frequently co-occurring words while the scanning window reads through the whole text. In communication research, these words are generally key words (or issues or symbols) represented in the text (Doerfel and Barnett, 1999; Hsu and Park, 2001). The default value of 25 is used in this research. In this research, the answers of each individual webmaster are separated from each other by a delimiter, which allows the analysis of each word in the text and in the context of the entire webmasters' answer, rather than as only single words.
Cluster Analysis
Since the sij contains numeric weights reflecting strengths of connections between words i and j, a number of different mathematical or statistical methods may be applied to S to facilitate the description of the semantic structure of the text. In this research, cluster analysis is employed to describe the meaning structure of the text. Cluster analysis identifies those groupings or clusters of nodes (in this case, words) that best represent their measured relations (Aldenderfer and Blashfield, 1984).
Ward's hierarchical clustering method (Ward, 1963) is used here. The Ward clustering method is agglomerative. Agglomerative clustering methods are methods that use a measure of distance between two objects and merge the closer ones (Griths et al., 1984). Simply speaking, it searches the pair of nodes with the greatest similarity (the smallest distance) from an n x n distance matrix of similarity (S). The process continues until there is only one object left. Ward's algorithm for agglomerative cluster analysis generally gives more sensible and interpretable solutions than other approaches if smaller concept groupings of texts are of interest.
The results of cluster analysis can be graphically displayed by CATPAC. CATPAC produces visual relationships among the words using a dendrogram. The dendrogram shows the pattern of relation between key words and further helps identify clusters of them. It also reveals where the clusters are divided and how strong the clusters are.
Results
Descriptive Statistics and Reliability
Seventy-one Webmasters participated in the survey. Seven of the remaining 71 were discarded because they contained unreliable answers. Sixty-four questionnaires were entered into SPSS for analysis (a response rate of 43 percent). Table 1 shows how the Web sites are distributed across the Yahoo categories.
Table 1: Classification of the sample
Category Web sites Number Arts and Humanities 0 Business and Economy samsungfire.com, samsungcard.co.kr, daishin.co.kr, hhi.co.kr, samsunglife.com, sklife.co.kr, ssyenc.co.kr, namkwang.co.kr, ssyapt.co.kr, samsung.com, nababa.tv, lotte.com, nalco.com, hdec.co.kr, kabc.or.kr, shinbo.co.kr, medidas.co.kr 17 Computers and Internet pricengo.com, bigfamily.co.kr, vpay.co.kr, mot.co.kr, auction.co.kr, sicc.co.kr, ntech9.com, kt.co.kr, hananet.net, shinbiro.com, intergump.co.kr, lghitachi.co.kr, icc.or.kr, edu.info21.or.kr, daum.net, popdesk.co.kr, webbbs.icc.or.kr, tiptech.net 18 Education kmu.ac.kr 1 Entertainment 0 Government kbe.go.kr, metro.ulsan.co.kr, andong.go.kr, pohang.kyongbuk.kr 4 Health healthkorea.net, med.yu.ac.kr, pohangsmh.co.kr, inha.com, akom.org, khmc.or.kr 6 News and Media joins.com, donga.com, hani.co.kr 3 Recreation and Sports koreanair.co.kr, weathernews.co.kr, chejushilla.com, woobangsp.co.kr 4 Reference 0 Regional council.chonnam.kr, council.daegu.go.kr, gncl.co.kr, council.provin.kyongbuk.kr, council.gwangju.kr, assem.cb21.net, council.pusan.kr, council.metro.incheon.kr, council.provin.chungnam.kr 9 Science 0 Social Science lif.or.kr 1 Society and Culture jubu21.net 1 Total 64
Next, all credibility items were analyzed. Missing values were excluded list-wise. Webmasters were asked to rate the items found in the following question: "When deciding on a hyperlink with other Web sites, how important do you think their credibility is in terms of truthfulness, reputation, usefulness, timeliness, competency, security, and reliability?" Remember that all items were measured using a fractionation scale. The question asked respondents to let 50 represent an average judgment of a given topic and zero represent none. As Table 2 shows, all seven items were reported to be more important than average. The scoring of items is as follows: truthfulness 91.56, reputation 78.91, usefulness 105.81, timeliness 85.94, competency 90.31, security 82.73, and reliability 79.84. The most important item was "usefulness." It is about twice as important as that of an average level. The next most important items were "truthfulness" and "competency." The difference among each category is not large. Respondents were also asked to judge the statement, "Overall, how much do you think your Web site's credibility increased when linking with other sites perceived highly credible?" The mean for this item was 77.58. Again, 50 represents the average judgment of a given topic. Thus, Webmasters surveyed answered that the perceived credibility of their Web sites would go up by 77.58 when hyperlinking with credible sites.
Table 2: Descriptive Statistics of Credibility ItemsN=64
Items Mean S.D. Minimum Maximum Truthfulness 91.56 33.24 50 150 Reputation 78.91 34.32 30 150 Usefulness 105.81 40.07 50 300 Timeliness 85.94 31.55 50 150 Competency 90.31 36.63 40 200 Security 82.73 37.70 30 150 Reliability 79.84 43.72 30 300 Overall 77.58 27.75 20 150
Factor Analysis
Factor analysis was done using the principal factors extraction method and a Varimax rotation. The criterion for retaining factors was a minimum Eigen value of 1.0. None of the cases had missing values.
The results showed that three underlying dimensions were found. They coincided with theoretical prediction. Two items (truthfulness and reputation) loaded on factor 1 (trustworthiness factor). Three items (usefulness, timeliness, and competency) loaded on factor 2 (expertness). Security and reliability loaded on Factor 3 (safety factor). The three dimensions account for 78.94 percent of the variances. The component matrix is summarized in Table 3.
Table 3: Factor Analysis of WebmastersReliability coefficient (Cronbach's Alpha) = .81
Component Loadings Items 1 2 3 Truthfulness 0.810 Reputation 0.761 Usefulness 0.721 Timeliness 0.845 Competency 0.890 Security 0.609 Reliability 0.951
Word Frequency Analysis
Webmasters were asked to give responses to the following open-ended questions: 1) a reason to choose a hyperlink with a certain Web site; and, 2) the advantage of hyperlinks. Computer content analysis using the CATPAC was done.
First, the results of question 1 showed that the most frequently mentioned word was "contents" which occurred 25 times by 20 (31.2 percent) Webmasters. The second most often occurring words were "relevance" and "similar" which were used 17 times by 17 (26.6 percent) organizations respectively. Their high frequency may be taken as an indicator of the validity of these procedures. The major purpose of a hyperlink between two Web sites (or Web pages) is to keep two interrelated sites (or Web documents) together. The next most frequently mentioned words were: "business," "type," "company," "information," "useful," "topic," and "advertising." Besides these ten words, terms such as "affiliation," "augmenting," "marketing," "banner," "mother," "sharing," "current," "exchange," "revenue," "win," "connectedness," "interface," "public," "quality," and "relations," frequently appeared.
Next, the results of question 2 showed that the most frequently mentioned word was "information" which occurred 14 times by 13 (20.3 percent) Webmasters. The second most often occurring word was "contents," which was used 13 times by 13 (20.3 percent) organizations. Their high frequency may be taken as an indicator of the validity of these procedures. A Web is a systematic body of Web sites that contains information. The information or contents are exchanged through the use of hyperlinks. The next most frequently mentioned words were: "augmenting," "navigation," "Web site," "current," "don't," "know," "relevant," and "convenience." Besides these ten words, terms such as "increasing," "interface," "marketing," "portal," "pursuing," "transfer," "various," "advertising," "exchange," "providing," "trust," "user," "kinds," "revenue," and "visitors" frequently appeared.
Cluster Analysis
Webmasters were asked to respond to the two following open-ended questions: 1) a reason to choose a hyperlink with a certain Web site; and, 2) the advantage of hyperlinks. Computer content analysis using the CATPAC was done.
First, the cluster analysis of question 1 resulted in one major grouping of words and six minor groups. The major cluster, the largest, included seven words. The members that consist of this cluster occurred 106 times (51.7 percent). The average number of Webmasters which mentioned members of this cluster was 14. Its members were: affiliation, relevance, topic, similar, business, type, and contents. This cluster reveals that most Web sites provide hyperlinks to their affiliated sites. The content similarity influences the affiliation among Web sites. Web sites choose their affiliation sites based on the type of businesses.
Six minor clusters are as follows. Cluster 2 (the first minor cluster), is comprised of three words: information, sharing, and useful. This cluster identifies those Webmasters whose answers concern the sharing of useful information among Web sites. Cluster 3 is also composed of three terms: marketing, public, and relations. This cluster suggests that the exchange of hyperlinks may contribute to increasing a marketing opportunity on the Web. The fourth cluster consists of the five following words: connectedness, interface, quality, exchange, and win. This cluster is a group that does not render a clear interpretation. Cluster 5, comprised of two members, includes current and augmenting. This shows that Web sites determine their hyperlinking to augment information on their Web sites. In cluster 6, mother and company frequently occurred together. This means that some Web sites systematically cooperate with other sites belonging to the same mother company. Those sites are usually called sister sites. Cluster 7, the last minor cluster, contains revenue, banner, and advertising. This shows that one revenue source for Web sites was advertisements placed on their Web pages.
Second, the cluster analysis of question 2 revealed that the 25 most frequently occurring words formed one major grouping and seven minor groups. Overall, the results of question 2 are quite similar to those of question 1. The major cluster, the largest cluster consisting of seven members, contains various, kinds, providing, convenience, relevant, information, exchange, and contents. These words occurred 54 times (38.3 percent). Eight Webmasters, on average, used these words in their answers. As described above, Webmasters expected the following aspects when deciding on a hyperlink with other Web sites: relevant information, content exchange, and augmenting current content.
Seven minor clusters are as follows. Three terms (revenue, marketing, and advertising) create the first minor cluster. This cluster reflects that a hyperlink is used as a new advertising vehicle. The second cluster, including current and augmenting, shows that Web sites offer links to other sites to make up for, or to expand, their own contents. There are two words, increasing and visitors, in the third minor cluster. Hyperlinking to other sites may lure Internet users to their own sites. In cluster 4, three terms related to Web site design (user, navigation, and interface) frequently occur together. A well-organized hyperlink structure on the Web can improve user-interface. The cluster 5 is made up of trust and transfer. Selecting a hyperlink to credible sites enhances an Internet user's trust. A sixth smaller cluster occurs with two words, don't and know. This cluster shows that Webmasters are not sure about the advantages of all hyperlinks. The last minor cluster is comprised of three terms: Web site, pursuing, and portal, This indicates that the advantage of hyperlinking is to provide a wide range of information.
In summary, the results of word frequency and clustering analysis suggest that hyperlinking motivations and advantages can be classified into two dimensions, navigational and for business purposes.
Discussion
The results of the survey reveal that Webmasters consider the credibility of hyperlinked Web sites important when deciding to hyperlink. All seven items that comprise Web site credibility were reported to be significant. Particularly, "usefulness" had the highest score, indicating that Web sites are more likely to hyperlink to Web sites having valuable and practical contents, information, or services. "Truthfulness" and "competency" followed in importance. In addition, a hyperlink to a credible site contributed to increasing a site's credibility.
The most frequently mentioned words and their clustering results indicate the motivations and advantages of hyperlinking with other sites. The results show that Webmasters choose hyperlinks with other sites to provide their visitors with a variety of relevant and useful contents; hyperlinking meets their expectations. Interestingly, word frequencies and clustering suggest that hyperlinking motivations and advantages are largely grouped into two dimensions, navigational functionality and business purposes.
Hyperlink navigability refers to a specific function that leads an Internet user to other Web sites via hyperlinks. The hyperlinked site is usually related to the hyperlinking site in terms of the type of information (e.g., a common topic) or service offered. In other words, hyperlinking is made to provide additional or background information or services. Links afford users the opportunity to find related Web sites so that the credibility of a hyperlinking Web site is accumulated through hyperlinks that point to credible sites. Previous research (Fogg et al., 2001; Nielsen, 2000; Rieh and Becklin, 1998) supported the notion that easy, clear, and accurate navigation based on the structure of hyperlinks has a positive impact on Web site credibility. Thelwall's (2001) study demonstrated that 61 sites - out of 232 commercial Web sites - had outgoing hyperlinks designed to enhance functionality.
The second motivational dimension that influences hyperlinking is business purpose. Webmasters said they tend to use hyperlinks to market a given Web site. For example, some Web sites hyperlink to other sites because they have a certain kind of business partnership. Thelwall (2001) demonstrated that business links were the most common kind of external link. Seventy-two sites out of 232 sites were found to have links based on any affiliated business relations. Particularly, hyperlink-based marketing among organizations appears to be popular (Hoffman and Novak, 1996; 2000). Hyperlink marketing refers to a marketing system by which goods and materials are sold and purchased, and advertisements are delivered via hyperlinks on the Web. For instance, consumers visit and navigate through a site. They click through a paid advertisement leading to other Web sites such as those providing online shopping. They may then buy products from the Web site. The hyperlinked Web site gives the hyperlinking site a percentage of its income.
From the viewpoint of task-focused or social-focused interorganizational communication (Haythornthwaite and Wellman, 1998; Kettinger and Grover, 1997), interorganizational hyperlink use is more likely to belong to inteorganizational relations for task purposes since hyperlinking tends to reflect the presence of business ties on the Web. However, given that computer-mediated communication patterns are influenced by social norms and practices (Hinds and Kiesler, 1995; Spears et al., in press), we should be careful to conclude that the hyperlinking among organizations are based on task-oriented interorganizational relations. The way in which organizations form cross-links keeps changing, depending on a variety of factors (DeSanctis and Monge, 1998; Mintzberg, 1973).
Webmasters also see the reliability of technical systems, such as online payments, as a major factor defining Web site credibility. This is consistent with past research. Park et al. (2002) analyzed the affiliation network among Korea's most frequently visited 152 Web sites. They found that the affiliation network was influenced by financial Web sites with which others were affiliated. They argued that having affiliations with financial Web sites, such as credit card companies, may be an effective way for common Web sites to reduce transaction risk and gain consumers' trust. In Korea, it is important to have an affiliation with a technically reliable Web site. The use of encrypted transactions on the Web is one of the determining factors impacting the spread of e-commerce activities. Among OECD countries, Korea's availability of secure servers affecting encrypted transactions is relatively low (OECD, 2001). Hagel and Armstrong (1997) also argue that, in the early stage of a virtual community, a specialized technical support service can play a principal role in organizing Web sites that are highly fragmented across multiple communities. These may be those sites with technical skills in operating online monetary transactions. Ng, et al. (1998) confirmed this observation by surveyed 98 companies listed on Yahoo.com. They found that 57.1 percent of the companies surveyed noted that the development of secure sites was the most important problem they had to solve for the success of e-commerce and 37.8 percent pointed out the need for suitable payment systems such as digital cash.
Limitations and Future Research
This research examined the determinants of hyperlinking among Web sites from the perspective of a Web site's credibility. However, the findings of this research should be viewed skeptically. Because the sample is comprised of only 64 Web sites, the Web sites in the sample may not be considered representative of other Web sites. In addition to a Web site's credibility, such factors as homophilous attributes (Park et al., 2000), interface usability (Park, 2001), and interpersonal communication may account for hyperlinking among Web sites. As Kim (2000) argues, hyperlinking may be an inherently multidimensional behavior influenced by a number of reasons. Also, the results describe the reports of Korean Webmasters. Despite South Korea's significance in relation to the development of the Internet, it may be a hasty generalization to regard results based upon Korean Web sites as a global phenomenon. Future research will investigate North American and European nations that play a more prominent role on the Internet. As data becomes accessible, this research will be extended to an international comparison.
About the Author
Han Woo Park is a research associate in the Networked Research and Digital Information group of the Royal Netherlands Academy (NIWI-KNAW, http://www.niwi.knaw.nl/nerdi) in Amsterdam. He recently completed his Ph.D. in the School of Informatics at the State University of New York at Buffalo. His research focuses on hyperlink network analysis and role of international and social factors in relation to new technologies.
E-mail: Han.Park@niwi.knaw.nl
Acknowledgments
The author is grateful to Tae Woo Park, In Yong Nam, Min Jeong Kim, Julia Hsu and George Barnett for their help while gathering the data for the article. Also, many thanks to Alexander Halavais for his insightful suggestions on my hyperlink analyses.
Notes
1. Rogers and Bhowmik, 1971, p. 209.
2. Source: Korean Network Information Center.
References
L.A. Adamic and E. Adar, 2001. "You are what you link," paper presented at the 10th annual International World Wide Web Conference, Hong Kong, at http://www10.org/program/society/yawyl/YouAreWhatYouLink.htm, accessed 19 June 2001.
R. Albert, H. Jeong and A.L. Barabási, 1999. "Diameter of the World Wide Web," Nature, volume 401, pp. 130-131.
M.S. Aldenderfer and R.K. Blashfield, 1984. Cluster Analysis. Beverly Hills, Calif.: Sage.
G.A. Barnett, B. Chon, H. Park and D. Rosen, 2001. "An Examination of International Internet Flows: An Autopoietic Model," paper presented at the annual conference of International Communication Association, Washington, D.C.,
G.A. Barnett, D.M. Hamlin and J.A. Danowski, 1981. "The use of fractionation scales in communication audits," Communication Yearbook, volume 5, pp. 455-471.
D. Berlo, J. Lemert and R. Mertz, 1969. "Dimensions for evaluating the acceptability of message sources". Public Opinion Quarterly, volume 33, number 4, pp. 563-576.
C.L. Borgman and J. Furner, 2001. "Scholarly communication and bibliometrics." In: B. Cronin and D. Shaw (editors). Annual Review of Information Science and Technology, volume 36, and at http://skipper.gseis.ucla.edu/faculty/jfurner/HTML/arist02.pdf.
R.W. Brislin, 1972. "Translation issues: Multilanguage versions and writing translatable English," American Psychologist, volume 7, pp. 299-300.
R.W. Brislin and W.H. Sinaiko, 1973. "Evaluating language translations: Experiments on three assessment methods," Journal of Philosophy and Rhetoric, volume 57, pp. 328-334.
A. Broder, R. Kumar, F. Maghoul, P. Raghavan, S. Rajagopalan, R. Stata, A. Tomkins and J. Wiener, 2000. "Graph structure in the Web," paper presented to the 9th International World Wide Web Conference, Amsterdam, at http://www9.org/w9cdrom/160/160.html, accessed 16 August 2001.
S.D. Brunn and M. Dodge, 2001. "Mapping the 'Worlds' of the World Wide Web: (Re)Structuring global commerce through hyperlinks," American Behavioral Scientist, volume 44, number 10, pp. 1717-1739.
J. Danowski, 1993. "Network analysis of message content," In: G. Barnett and W. Richards, (editors). Progress in Communication Sciences, Norwood, N.J.: Ablex, pp. 197-222.
G. DeSanctis and P. Monge, 1998. "Communication processes for virtual organizations," Journal of Computer-Mediated Communication, volume 3, number 4, and at http://www.ascusc.org/jcmc/vol3/issue4/desanctis.html, accessed 19 September 2000.
M.L. Doerfel and G. Barnett, 1999. "A semantic network analysis of the International Communication Association," Human Communication Research, volume 25, number 4, pp. 589-603.
S.M. Ervin and R.F. Bowers, 1952. "Translation problems in international surveys," Public Opinion Quarterly, volume 16, pp. 595-604.
S.M. Ervin and C.E. Osgood, 1954. "Second language learning and bilingualism," Journal of Abnormal Psychology, volume 49, pp. 139-146.
P. Evans and T.S. Wurster, 1999. "Getting real about virtual commerce: Strategies for success in electronic and physical retail commerce," Harvard Business Review,, volume 77, number 6 (November-December), pp. 84-94.
A.J. Flanagin and M.J. Metzger, 2000. "Perceptions of Internet information credibility," Journalism and Mass Communication Quarterly, volume 77, number 3, pp. 515-540.
B.J. Fogg, J. Marshall, O. Laraki, A. Osipovich, C. Varma, N. Fang, J. Paul, A. Rangnekar, J. Shon, P. Swani and M. Treinen, 2001. "What makes Web sites credible? A report on a large quantitative study," paper presented at the Computer-Human Interaction Conference, Seattle, Washington.
D. Gefen, 2000. "E-commerce: The role of familiarity and trust," OMEGA-International Journal of Management Science, volume 28, number 6, pp. 725-737.
U. Gretzel, 2001. "Internet readiness among tourism organizations: An empirical test of a stages model of technology adoption and use," paper presented at the Organizational Communication Mini-Conference, Champaign, Ill.
A. Griths, L.A. Robinson and P. Willett, 1984. "Hierarchic agglomerative clustering methods for automatic document classification," Journal of Documentation, volume 40, pp. 175-205.
J. Hagel, III and A.G. Armstrong, 1997. Net Gain: Expanding Markets Through Virtual Communities. Boston: Harvard Business School Press.
A. Halavais, 2000. "National borders on the World Wide Web," New Media and Society,, volume 2, number 1, pp. 7-28.
B. Hayes, 2000. "Computing science graph theory in practice: Part 1," American Scientist, volume 88, number 1, pp. 9-13.
C. Haythornthwaite and B. Wellman, 1998. "Work, friendship and media use for information exchange in a networked organization," Journal of the American Society for Information Science, volume 46, number 12, pp. 1101-1114.
M.R. Henzinger, 2001. "Hyperlink analysis for the Web," IEEE Internet Computing, volume 5, number 1, pp. 45-50.
P. Hinds and S. Kiesler, 1995. "Communication across boundaries: Work, structure, and use of communication technologies in a large organization," Organization Science, volume 6, number 4, pp. 373-393.
D.L. Hoffman and T.P. Novak, 2000. "How to acquire customers on the Web," Harvard Business Review, volume 78, number 3, pp. 179-188.
D.L. Hoffman and T.P. Novak, 1996. "Marketing in hypermedia computer-mediated environments: Conceptual foundations," Journal of Marketing, volume 60, number 3, pp. 50-68.
C. Hovland, I. Janis and H. Kelley, 1953. Communication and Persuasion. New Haven, Conn.: Yale University Press.
C. Hovland and W. Weiss, 1951. "The influence of source credibility on communication effectiveness," Public Opinion Quarterly, volume 15, number 4, pp. 633-650.
C.W. Hsu and H.W. Park, 2001. "An identification of online privacy issues - Content analysis of top 500 Web sites," paper presented at the Second International Conference of the Association of Internet Researchers, Minneapolis.
P. Ingwersen, 1998. "The calculation of Web impact factors," Journal of Documentation, volume 54, number 2, pp. 236-243.
M.H. Jackson, 1997. "Assessing the structure of communication on the World Wide Web," Journal of Computer-Mediated Communication, volume 3, number 1, at http://www.ascusc.org/jcmc/vol3/issue1/jackson.html, accessed 19 September 2000.
T. Johnson and B. Kaye, 1998. "Cruising is believing?: Comparing the internet and traditional sources on media credibility measures," Journalism and Mass Communication Quarterly, volume 75, number 2, pp. 325-340.
W. Kettinger and V. Grover, 1997. "The use of computer-mediated communication in an interorganizational context," Decision Sciences, volume 28, number 3, pp. 513-555.
H.J. Kim, 2000. "Motivations for hyper-linking in scholarly electronic articles: A qualitative study," Journal of the American Society for Information Science, volume 51, number 10, pp. 887-899.
S. Kiousis, 1999. "Public trust or mistrust?: Perceptions of media credibility in the information age," paper presented at the annual conference of the Association in Education for Journalism and Mass Communication, New Orleans.
J.M. Kleinberg, 1999a. "Hubs, authorities, and communities," ACM Computing Surveys, volume 31, number 4, at http://www.cs.brown.edu/memex/ACM_HypertextTestbed/papers/10.htm, accessed 2 March 2001.
J.M. Kleinberg, 1999b. "Authoritative sources in a hyperlinked environment," Journal of the ACM, volume 46, number 5, pp. 604-632.
Korea Herald, 2001. "Broadband penetration nears saturation in Korea," at http://www.koreaherald.co.kr/SITE/data/html_dir/2001/09/27/200109270007.asp, accessed 27 September 2001.
V. Kreb, 2000. "Working in the connected world book network," IHRIM (International Association for Human Resource Information Management) Journal, volume 4, number 1, pp. 87-90.
K. Krippendorff, 1989. "Content analysis: An introduction to its methodology," Sage Commtext Series, volume 5, pp. 12-157.
J.C. McCroskey and J.J. Teven, 1999. "Goodwill: A reexamination of construct and its measurement." Communication Monographs, volume 66, pp. 90-103.
S. Milgram, 1967. "The small world problem," Psychology Today, volume 1, number 1, pp. 60-67.
H. Mintzberg, 1973. The Nature of Managerial Work. New York: Harper and Row.
H.-I. Ng, Y.J. Pan and T.D. Wilson, 1998. "Business Use of The World Wide Web: a report on further investigations," Information Research, volume 3, number 4, at http://informationr.net/ir/3-4/paper46.html, accessed 11 September 2001.
J. Nielsen, 2000. Designing Web usability. Indianapolis: New Riders.
Nielsen/NetRatings, 2001. "Korea Claims 23 of World's 100 Biggest Locally Accessed Web Properties." at http://www.nielsen-netratings.com/news.jsp, accessed 27 September 2001.
OECD (Organization for Economic Co-operation and Development), 2001. "Business-to-consumer e-commerce statistics," paper presented at the Consumers in the online marketplace: OECD workshop on the guidelines: One year later, Berlin.
OECD, 1998. Working paper on telecommunication and information service policies: Internet infrastructure indicators. Paris: OECD.
J.W. Palmer, J.P. Bailey and S. Faraj, 2000. "The role of intermediaries in the development of trust on the WWW: The use and prominence of trusted third parties and privacy statements," Journal of Computer-Mediated Communication, volume 5, number 3, at http://www.ascusc.org/jcmc/vol5/issue3/palmer.htm, accessed 22 June 2000.
H.W. Park, in press. "What is hyperlink network analysis?: New method for the study of social structure on the Web," Connections.
H.W. Park, 2001. "Older adults and persuasive technology - Developing persuasive principles and integrating into artifacts," paper presented at the annual conference of the International Communication Association, Washington, D.C.
H.W. Park, G.A. Barnett and I.Y. Nam, 2002. "Hyperlink-affiliation network structure of top Web sites: Examining affiliates with hyperlink in Korea," Journal of the American Society of Information Science and Technology, volume 53, number 7, pp. 602-611.
H.W. Park, G.A. Barnett and C.S. Kim, 2000. "Political communication structure in Internet networks - A Korean case," Sungkok Journalism Review, volume 11, pp. 67-89.
S.Y. Rieh and N.J. Becklin, 1998. "Understanding judgment of information quality and cognitive authority in WWW," Journal of the American Society for Information Science, volume 35, pp. 279-289.
E.M. Rogers and D.K. Bhowmik, 1971. "Homophily-heterophily: Relational concepts for communication research," In: L.L. Barker and R.J. Kibler (editors). Speech communication behavior: Perspectives and principles. Englewood Cliffs, N.J.: Prentice-Hall, pp. 206-225.
W. Schweiger, 2000. "Media credibility - Experience or image? A survey on the credibility of the World Wide Web in Germany in comparison to other media," European Journal of Communication, volume 15, number 1, pp. 37-59.
C. Shapiro and H.R. Varian, 1999. Information rules: A strategic guide to the network economy. Boston: Harvard Business School Press.
K. Sheehan, 2001. "E-mail survey response rates: A review," Journal of Computer-Mediated Communication, volume 6, number 2, at http://www.ascusc.org/jcmc/vol6/issue2/sheehan.htm, accessed 1 April 2001.
R. Spears, M. Lea and T. Postmes, in press. "Social psychological theories of computer mediated communication: Social pain or social gain?" In: P. Robinson and H. Giles, (editors). The handbook of language and social psychology. Chichester: Wiley, and at http://www.psy.uva.nl/ResEdu/SP/Research/PDF/hb_ch_rs.PDF, accessed 11 March 2002.
S.S. Sundar and C. Nass, 2001. "Conceptualizing sources in online news," Journal of Communication, volume 51, number 1, pp. 52-72.
L. Terveen and W. Hill, 1998. "Evaluating emergent collaboration on the Web," paper presented to the Conference on Computer Supported Cooperative Work, Seattle.
M. Thelwall, 2001. "Commercial Web site links," Internet Research: Electronic Networking Applications and Policy, volume 11, number 2, pp. 114-124.
S. Tseng and B. Fogg, 1999. "Credibility and computing technology," Communications of the ACM, volume 42, number 5, pp. 39-44.
J. Ward, 1963. "Hierarchical grouping to optimize an objective function," Journal of American Statistics Association, volume 37, pp. 236-244.
D.J. Watts and S.H. Strogatz, 1998. "Collective dynamics of 'small-world' networks," Nature, volume 393, pp. 440-442.
J. Woelfel, 1998. Catpac User's Manual. (Catpac II, Version 2.0). New York: Rah Press.
J. Woelfel and E.L. Fink, 1980. The Measurement of Communication Processes: Galileo Theory and Method. New York: Academic Press.
Editorial history
Paper received 29 April 2002; accepted 25 October 2002.
Copyright ©2002, First Monday
Copyright ©2002, Han Woo Park
Examining the Determinants of Who is Hyperlinked to Whom: A Survey of Webmasters in Korea by Han Woo Park
First Monday, volume 7, number 11 (November 2002),
URL: http://firstmonday.org/issues/issue7_11/park/index.html
