How to social scientists use link data (11 june2010)

WCUWEBOMETRICSINSTITUTEINVESTIGATING INTERNET-BASED POLITICSS WITH E-RESEARCH TOOLS

Use of search engine data to understand Internet-based political and electoral communication?

Dr. Han Woo PARK

Associate ProfessorDepartment of Media & CommunicationYeungNam University214-1 Dae-dong, Gyeongsan-si, Gyeongsangbuk-do 712-749Republic of [email protected]://www.hanpark.net

Principal Investigator of WCU Project: Investigating Internet-based Politics with e-Research Tools.

A paper accepted for presentation to the 2010 annual conference of ICA (International Communication Association) held in Singapore, http://www.ica2010.sg

http://yu.ac.kr/~hanpark/HanWooParkVita.PDF

http://www.yu.ac.kr/_english/main/index.php

http://www.hanpark.net/

http://www.ica2010.sg/

WCUWEBOMETRICSINSTITUTEINVESTIGATING INTERNET-BASED POLITICS WITH E-RESEARCH TOOLS

AbstractAs link data become more available from search engines, the data can play an important role in understanding socio-political activities in cyberspace.

Two case

Internet-based political and electoral communications conducted in Korea, where online communication is among the most well-established in the world.

link data collected from search engines have several merits for political communication research, including inexpensive access to web resources and ease of use.

link data have some weakness, such as reliability issues: internal algorithms are not made in public and there are inconsistencies among search engines.

Two case studies to draw conclusions about the use of link data drawn from search engines for research purposes.

Advantage Disadvantage

Key worlds: search engine, link analysis, social science method, political communication, online research method


IntroductionWeb traffic is only one part of online communication. There are other significant communication behaviors on the Internet that could be captured by researchers using publicly available tools. The Myth of Digital Democracy, Hindman (2008)

The Internet is growing at an enormous rate in terms of recorded digital traces.

political and electoral communication is undergoing a significant transformation because Internet-mediated communication has rapidly become popular among government officials, politicians, and citizens.

These bits of online information produced and inscribed via links are explicitly or implicitly connected to each other by topics of interest and have become a useful resource for researchers.

In this study, we are mostly interested in examining and evaluating the practices and trends of online political communication by examining the ‘linking’ behavior involved.


IntroductionThe Internet represents a massive storehouse of social networking data, enabling researchers to capture political communications by individual officials, politicians, and activists. This is particularly true in South Korea, where political participants almost always have online representation.

Two case studies of Internet-based political and electoral communications conducted in Korea.

Examines the online characteristics of the 2007 presidential election in Korea.We analyze a substantial amount of link data obtained using search engines in order to understand their broader political implications.

Case 1.

Case 2.

Examines how search engine-based methods are useful in discovering connections among Korean legislators in the online social networking sphere in Korea.

These cases reveal that link data collected from search engines have merits for political communication research. The following discussion looks at the pros and cons of search engines as research tools and the use of link data collected from search engines.


Background Information and Relevant Literature

The web is constantly growing in terms of content, search engines play the role of information intermediaries for the web, which has become an important part of public dialogue.

Thelwall (2004) Search engines allow researchers to get materials from a much wider range of subjects than previously available through traditional library databases. One important contribution of search engines is the ability to retrieve relevant web-based materials using hypertext links, enabling researchers to gather and recombine information collected from the Internet.

In research analyzing the e-science landscape using statistics about individual keyword queries via Google, Meyer and Schroeder (2009) found that the search pattern for the term ‘cyberinfrastructure’ reveals which U.S. regions are deeply involved in national e-science projects, such as major supercomputing centers. Further, using Google-generated keyword activity data, Varian and Choi (2009) showed that public data from search engines helped improve economic forecasts for various industries, including automobile sales, home sales, retail sales, and travel behavior.


Background Information and Relevant Literature

Elgesem (2008, p.239) argued that “search engines are objective in the sense that these engines try to be consistent with their own stated policies.” With respect to coverage and consistency, problems may occur due to the nature of the unstructured web. In other words, the lack of reliability may not be caused by the search engine.

Science organizes, structures, and evaluates information to develop a systematic body of knowledge. It is up to the researcher to draw the appropriate conclusions, using his or her expertise, about the information gathered from the web using search engines. While search engines collect data from the entire web, finding the truth from the information is the business of academics (Caldas et al., 2008).


Case 1. 2007 Korean Presidential ElectionBackground

Korea boasts the highest proportion of broadband users in the world, and there is a unique evolution of online culture in Korean cyberspace. The country’s impressive level of technological development includes a vibrant online communication environment.

The online political climate during the 2007 Korean presidential election can be examined effectively using web-based data analysis. In particular, there were 12 candidates who ran for president and several parties were created in 2007 to support these candidates (see Table 1). The candidates and the parties had to compete against each other to win public attention, particular since it was difficult for citizens to differentiate their stances on issues.

Particularly useful for web analysis was the fact that public opinion surveys could not be published within six days before the 2007 election. In 2002, surveys could not be published within 22 days of the presidential election. We will examine how the popularity of individual candidates and parties developed during the 2007 presidential election campaign in South Korea using web-based data collection.


Background

Case 1. 2007 Korean Presidential Election

Table 1. Websites of Presidential Candidates and Parties

Candidate Website Candidate’s Party WebsiteLee Myung-Bak (Lee MB) www.mbplaza.net Grand National Party (GNP) www.hannara.or.kr

Chung Dong-Young (Chung DY) www.cdy21.net United New Democratic Party (UNDP) www.undp.kr

Lee Hoi-Chang (Lee HC) www.leehc.org Independent

Moon Kook-Hyun (Moon KH) www.moon21.kr Creative Party (CKP) www.ckp.kr

Kwon Young-Ghil (Kwon YG) www.ghil.net Democratic Labor Party (DLP) www.kdlp.org

Rhee In-Je (Rhee IJ) www.ijworld.or.kr Democratic Party (DP) www.minjoo.or.kr

Huh Kyung-Young (Huh KY) Same as party site Economy & Republican Party (ERP) www.gonghwa.com

Geum Min (Geum M) www.minnmin.net Socialist Party (KSP) www.sp.or.kr

Chung Kun-Mo (Chung KM) www.bestjung.kr True Owner Coalition (TOC) www.chamjuin.or.kr

Chun Kwan (Chun K) www.chamsaram.or.kr Chamsaram Society Full True Act (CSFTA)

Same as candidate site

Sim Dae-Pyeng (Sim DY) www.dpsim.co.kr People First Party (PFP) www.mypfp.or.kr

Lee Soo-Sung (Lee SS) www.leesoosung.com People’s Coalition (PC) Same as candidate site



Data collection

Subject investigated: The web presence of all 12 candidates and their parties during the official campaign period of the 2007 Korean presidential election.

Date: The online visibility of each party and candidate was measured with colink data, which refer to incoming links shared between any pair of candidates’ or parties’ websites. These data were collected at multiple points during the election period from search engine indexes.

Acquisition time:Colink data were gathered in three-day intervals, 29 November - 17 December in 2007 (the election was held in 19 December 2007).



Analytic technique

The process of generating a colink matrix is as follows (for detailed procedures, see Thelwall, 2009).

First, an input file, which lists a set of colink queries to retrieve webpages sending at least one link to a pair of website addresses under investigation, is made using a text editor program (e.g., NotePad).

Second, a symmetrical website*website matrix is created.

Third, the resultant sociogram reveals which sites are central and peripheral.

Fourth, key social network measures are calculated to analyze the overall structure of colinks: clustering coefficient, geodesic distance, and degree centrality (For the detailed explanation of network metrics, see Wasserman & Faust, 1994; Park & Leydesdorff, 2009).



As Park (2003) stated in his seminal article titled “Hyperlink network analysis: A new method for the study of social structure on the web,” link data can be examined using social network analysis.

The clustering coefficient indicates the degree to which friends (in this case, neighboring websites) of a person know each other; average distance value shows the degree of separation on average between each (reachable) pair of websites in terms of geodesic distance that is defined as the shortest path. The average distance is often reported with an overall cohesion value between 0 and 1. Larger values indicate greater cohesiveness.

Degree centrality is a basic and primary measure to indicate the position of a website based on its interconnections with other websites in a given network. Due to space limitations, only three websites used in this study were reported in Table 2.

Analytic technique



Findings

The results from the colink data and its accompanying qualitative investigation indicate the following. The hyperlinked relations among the candidates and parties over time reflect their positioning in the electoral race and the general climate of offline politics.

In Figure 1, squares represent candidates’ sites and circles represent parties’ sites. The thickness of the lines connecting them is drawn bigger in accordance with the number of external colinks directed to a pair of sites.



Findings

Figure 1: 2 Dec 2007

Since the snapshots taken in December 2007 display distinctive changes, we focus on these dates in the results section. On 2 December, we can see clearly that parties are clustered with other parties and candidates are closely linked with other candidates. However, 9 out of 20 sites were not included in the network. Isolated candidate and party sites are located in the upper left-hand corner of the diagram. During the initial stage of the campaign, online attention focused on major parties and candidates, as Figure 1 shows. Overall, the network structures among candidates and parties seemed to be loose.



FindingsFrom the network diagram captured on 11 December, it is clear that the Lee HC (leehc.org) website was successful in catching up with other major candidates’ websites. As a matter of fact, he became one of the major candidates. He took third position in the presidential race by gaining 15.1% of the total votes. But, he entered the race late relative to other major candidates and ran without support from a specific party. As seen in Figure 2, he was finally included in the user-driven networking zone. Another interesting finding on 11 December is the position of ijworld.or.kr (Rhee IJ’s site) and ckp.kr (CKP’s site). In the 2 December figure, ijworld.or.kr was only connected to the party site (minjoo.or.kr), but it later became linked with progressive candidate Chung DY’s site (cdy21.net). The reason is that Rhee IJ had been strongly urged by the public to give up his candidacy in support of Chung DY. In the case of ckp.kr, this website showed a dramatic turnaround as its candidate’s (Moon KH, moon21.kr) popularity increased.




Findings

Figure 3, captured on 17 December, shows that, at the end of the campaign period, a total of 14 candidates and parties joined the user-driven networks. As election day approached, the number of isolated sites noticeably decreased from 9 (2 Dec) to 5 (17 Dec), which reflected increasing online public awareness of the election. Two candidates, Lee HC (leehc.org) and Rhee IJ (ijworld.or.kr), seemed to be on their way to success because they were fully included in the network. President-elect Lee MB (mbplaza.net ) and his party (hanara.or.kr) were in the center of the candidate and party group networks, respectively. Overall, the figures show that as time progressed, link density increased.




FindingsIn addition to the three figures, other network measures between nodes were examined, as summarized in Table 2. The clustering coefficient that measures the cohesiveness of the network was 2.581 in the beginning of the official campaign (2 Dec), but dropped to 1.777 two days before the election (17 Dec). Next, while average reachability scores in terms of geodesic distance between pairs show some fluctuation, cohesion values in parenthesis clearly indicate that the network became dense (0.215 -> 0.273 -> 0.346). Average distance values lengthened in accordance with the gradual influx of isolated sites into the network. Lastly, three websites in particular showed an increase in their degree centralities.

Table 2. Network Measures of Colink Networks with Three Different Points

Network measures 2 Dec 2007 11 Dec 2007 17 Dec 2007

Clustering coefficient 2.581 2.368 1.777

Average distance

(Cohesion value)

1.564

(0.215)

1.821

(0.273)

1.681

(0.346)

Degree centralities of sites

ijworld.or.kr

leehc.org

ckp.kr

0.158

0.000

0.000

0.263

0.053

0.053

0.684

0.263

0.053

Note: Data were dichotomized for the calculation of clustering coefficient and geodesic distance values, and degree centralities were normalized for comparison across networks.


Case 2. Cyworld Mini-hompies of Korean Legislators

Case 2. Cyworld Mini-hompies of Korean LegislatorsBackground

The sample for the social networking sphere (Figure 4) is drawn from the Cyworld mini-hompies (mini-homepages) maintained by members of the 18th National Assembly. Cyworld is a Korean web-based social networking service launched on 1 September 1999 .“Cy” means both “cyber” in English and “relationship” in Korean. Internet users form a buddy relationship with each other through a service called “mini-hompy,” where they can decorate their “home” and invite friends using a photo gallery, video clip, message board, guestbook, friend list, and personal bulletin board. It is similar to the U.S.-based Facebook and MySpace sites. In a study of blogs of Korean legislators, Park and Kluver (2009) stated that the mini-hompies are not only popular with Korean youth, but also with many older Koreans. Cyworld is Korea’s dominant social networking site, with 90% of young people in their 20s and 25% of the total South Korean population registered as users. According to the study, South Korean politicians have also begun to make significant use of blogs and mini-hompies to expand their “personal communities.”



Data collection

We were interested in the legislators’ interpersonal social networking in cyberspace. However, it is difficult to trace personal relationship among politicians because “friends” on mini-hompies are confidential.

As of 16 March 2009, 83 out of a total of 295 National Assembly members run their own mini-hompies. The distribution of the 83 members is as follows: 47 from the GNP (Grand National Party, the ruling party, Han-na-ra in Korean), 19 from the DP (Democratic Party, the major opposition party, Min-joo in Korean), 8 from the LFP (Liberty Forward Party, Jayu Seonjindang in Korean), 2 from the PPA (Pro-Park Alliance, Chinbak Yeondae in Korean), 2 from the DLP (the Democratic Labor Party, progressive party, Min-joo-no-dong in Korean), 1 from the CKP (Creative Korea Party, Changjo Hangukdang), and 4 independents. We obtained the members’ mini-hompy URLs through official websites and commercial search engines.

Three datasets with respect to the status of legislators’ mini-hompies were collected: 1) date of creation, 2) number of visitors, and 3) frequency of bookmarking by other mini-hompy users. These data were manually collected. Legislators who did not indicate the creation date of their sites public were excluded. As a result, we collected data on links, creation date, number of visitors, and bookmarks for 70 Korean politicians.



The same procedures described earlier have been applied to this case study. Regarding social network analysis, betweenness centralities were measured in addition to degree centralities. While degree centrality measures a relational attribute of an actor in a network (e.g., popularity, activity, and prominence in the flow of information), betweenness is a useful index to measure the potential influence of the actor in controlling and coordinating communication. Additionally, we have examined the association between network centralities and traffic activities.

Analytic technique



Thirty out of 70 Cyworld-connected legislators were included in the colink network; 40 politicians operated isolated websites.

Findings

Lines between the websites indicate the presence of colinks between mini-hompies, the thickness of lines indicates the number of colinks, and colors represent the different parties. The size of the label (that is, the politician’s web address) is in proportion to the number of bookmarks, and the size of the node corresponds to the number of visitors. The map clearly distinguishes relatively peripheral groups (sparely linked politicians) from central ones (densely connected politicians) in Korea.



Findings

The network centralities and traffic information for the 30 Assembly members’ mini-hompies are presented in Table 3.

Notes: http://www.cyworld.com was omitted in front of the URLs in the table. Visitors and bookmarks were divided by the number of days the Cyworld mini-hompies were in existence.

Table 3. Normalized Measures of Individual PoliticiansName Cyworld_URL Degree Betweenness Visitors Bookmarks Party

Kang GG kanggigap 28.986 0.005 119.36 0.21 DLPKwon YG kwondlp 27.536 0 92.87 0.08 DLPKim MS kimmoosung 14.493 0 4.98 0.00 GNPKim YS kimyoungsun1 42.029 3.649 19.76 0.02 GNPKim UD kimeuldong 26.087 0 88.08 0.13 PPAKim JP jinpyo415 28.986 0.005 41.04 0.04 DPNa KY KyoungOne 28.986 0.005 502.42 0.44 GNPKP npil2580 28.986 0.005 32.32 0.05 GNPPark GH ghism 42.029 3.649 4524.34 5.89 GNPSeo BS sbs8680 14.493 0 3.91 0.00 GNPSong YS peacekeeping 14.493 0 28.33 0.02 PPAAhn HJ masanjun 14.493 0 7.35 0.00 GNPWon HY whywon21 28.986 0.005 9.28 0.01 DPWon HR wworld 28.986 0.005 86.71 0.17 GNPYoo SM ysm21com 14.493 0 11.70 0.01 GNPYoo JB ilovegimpo 14.493 0 6.47 0.00 GNPLee KJ yeskj1 28.986 0.005 15.08 0.02 DPRhee IJ ij21 28.986 0.005 17.96 0.02 IndyLee HC chang2007 27.536 0 2726.29 1.34 LFPChung MJ globalmj 28.986 0.005 87.14 0.19 GNPChung Sk skchung21 28.986 0.005 23.07 0.03 DPChung JS jangseon 28.986 0.005 6.55 0.00 DPCho KT boss110 28.986 0.005 138.22 0.13 DPJoo SY doitnow304 14.493 0 59.64 0.03 GNPChin SH sheechin 28.986 0.005 19.26 0.02 GNPCha MJ sosacha 28.986 0.005 26.40 0.01 GNPHan SK hsunkyo 14.493 0 53.54 0.11 GNPHong JP jphong21 28.986 0.005 61.89 0.05 GNPHwang WY hwangwygrace 28.986 0.005 8.44 0.01 GNPHwang JH uncinc 14.493 0 7.50 0.00 GNP

http://www.cyworld.com/

http://www.cyworld.com/

http://www.cyworld.com/kanggigap/

http://www.cyworld.com/kwondlp

http://www.cyworld.com/kimmoosung

http://www.cyworld.com/kimyoungsun1

http://www.cyworld.com/jinpyo415

http://www.cyworld.com/KyoungOne/

http://www.cyworld.com/npil2580

http://www.cyworld.com/ghism

http://www.cyworld.com/sbs8680

http://www.cyworld.com/peacekeeping

http://www.cyworld.com/masanjun

http://www.cyworld.com/whywon21/

http://www.cyworld.com/wworld

http://www.cyworld.com/ysm21com

http://www.cyworld.com/ilovegimpo/

http://www.cyworld.com/yeskj1

http://www.cyworld.com/ij21

http://www.cyworld.com/chang2007

http://www.cyworld.com/globalmj

http://www.cyworld.com/skchung21

http://www.cyworld.com/jangseon

http://www.cyworld.com/boss110/

http://www.cyworld.com/doitnow304

http://www.cyworld.com/sheechin

http://www.cyworld.com/sosacha



Figure 4: Cyworld Mini-hompies of Korean legislators

Findings

As seen in Figure 4, the network structure shows a clear butterfly pattern. There is one hub (ghism) that belongs to Park Gyun-Hye (Park GH, www.cyworld.com/ghism), the daughter of ex-president Park Jeong-Hee and one of two major GNP candidates (along with president-elect Lee MB) in the 2007 presidential race.



Findings

Table 4 demonstrates that the magnitude and direction of the association between several centralities and two other variables—visitors and bookmarks—are significantly correlated.

Table 4. Pearson’s coefficient

Betweenness Visitors BookmarksDegree .572(**) .390(*) .420(*)

Betweenness .572(**) .665(**)

Visitors .945(**)

** Correlation is significant at the 0.01 level (1-tailed).* Correlation is significant at the 0.05 level (1-tailed).N=30 (Isolated politicians were excluded in this analysis)


This analysis of the linking patterns among Korean politicians’ websites reveals some interesting facets of how search engine data can be used to map Korean Internet-mediated political culture. Our findings demonstrate that online markers of political communication, in the form of hyperlinks, exhibit the significant characteristics that drive many other phenomena in Korean presidential elections: the networked practices of politics and affiliations among politicians.

Discussion and Conclusion

Questions that require further exploration include:

1)Can we trust available search engines, and, if not, what can we do about them? If we are able to gain access to the search process and how websites are ranked, can this problem be solved?

2) How do we deal methodologically with the complexity of Web 2.0 using current search engine mechanisms that do not properly index multimedia content? Do we need new search tools for Web 2.0?



What are the lessons from this particular exercise for mapping and tracking online political communication in the future?

Lessons for Future Tracking and Mapping Research

First, some forms of political information exchange are difficult to study only using search engine results.

Second, qualitative analysis is also necessary because every country has different web characteristics.

Third, although we analyzed link data using social network measures and visualizations, this study is descriptive and topological.

Fourth, the purpose of this article was to introduce social science readers to the methods of search engine-mediated data collection for political research using social network measures and visualizations.


THE END

•Park, H. W. (2011 forthcoming). How do social scientists use link data from search engines to understand Internet-based political and electoral communication. Quality & Quantity

•The author would like to thank Bernie Horgan, Rob Ackland, Jeong-Soo Seo, and Yeon-ok Lee for their helpful comments on an earlier draft. Part of this research was carried out during the author’s stay at the Oxford Internet Institute. •During the preparation of final manuscript, this research is supported from the WCU project granted from South Korean Government. (No. 515-82-06574). http://english-webometrics.yu.ac.kr

Thank you for listening, and thank my assistants (Ting Wang)

How to social scientists use link data (11 june2010)

Education

consof search engines

online communication

search pattern

public data

search enginebased methods

enormous internet

use of link data

easesearch engines