![Sample our Communication Studies journals, sign in here to start your access, 2013 & 2014 volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5927/TOC-Subject-Sample-2021-Information-Science2.jpg"})
Abstract
This study examines information diffusion and the follower network among a group of Sina Weibo users interested in homeowner associations. Using social network analysis techniques, this paper explores the network structure, the formation of follower relations and information diffusion. It reveals that micro-blogging is an important online platform because it can conveniently and inexpensively foster public online issue-networks beyond geographical boundaries. Specifically, Weibo has the potential to enable cross-province networking and communication, although geographical proximity is still at work; the trustworthiness of micro-blog users indirectly contributes to information diffusion by facilitating the formation of follower relations; and issue-specific follower networks facilitate information diffusion pertinent to the issue at stake. These findings suggest that micro-blogging services might have long-term effects on collective action by fostering issue-networks among civil society organizations or activists in different provinces.
Acknowledgements
This study was sponsored by Chinese National Social Science Foundation (12CSH043) and Shanghai Pujiang Program (13PJC011). The authors thank Ngai-ming Yip, Wanxin Li and the anonymous reviewers for their valuable comments and suggestions.
Notes on contributors
Ronggui Huang is an Assistant Professor in the Department of Sociology, Fudan University. His research interests include Internet and contentious politics, neighborhood governance in China, social capital and social networks. [email: [email protected]]
Xiaoyi Sun is a Ph.D. candidate in the Department of Public Policy, City University of Hong Kong. Her research interests include urban studies and neighborhood governance in China. [email: [email protected]]
Notes
1. Collective action is defined as any action taken by two or more people in order to improve homeowners’ conditions.
2. http://news.sohu.com/20070126/n247846747.shtml, accessed on 1 July 2012.
3. http://citynews.eastday.com/csdb/html/2011-01/13/content_35673.htm, accessed on 1 July 2012.
4. http://www.sywy.net.cn/, accessed on 1 August 2012.
5. http://wzwx.66wz.com/system/2012/05/14/103166040.shtml, accessed on 1 August 2012.
6. http://jsnews.jschina.com.cn/system/2011/12/19/012324855.shtml, accessed on 1 August 2012.
7. http://news.cnwest.com/content/2012-10/14/content_7421913.htm, accessed on 1 August 2012.
8. http://www.zj.xinhuanet.com/video/2013-01/11/c_114337566.htm, accessed on 1 August 2012.
9. http://zjnews.zjol.com.cn/05zjnews/system/2012/12/04/018990587.shtml, accessed on 1 August 2012.
10. http://news.xinhuanet.com/tech/2012-02/29/c_122769084.htm, accessed on 1 August 2012.
11. Some argue that Twitter networks are not social networks because they exhibit low reciprocity and highly skewed degree distribution (Wu et al., Citation2011). On the contrary, others show that levels of reciprocity are moderate or high, and that levels of reciprocity increase when confined to users from one continent (Java et al., Citation2007). Zhang and Pentina (Citation2012) reviewed the previous literature on Twitter and concluded that Twitter is used to satisfy the needs of both information and social connection. As revealed in the empirical section, when indegree is confined to a group of users with shared interest, the follower network exhibits a relatively high level of reciprocity and is much closer to social networks. Therefore, the general theory of social networks can still shed some light on follower relation formation.
12. This study does not claim to exhaust all possible factors. First, ‘crawling’ data directly from a micro-blogging site is constrained by what exists on that site, a situation not unlike secondary data analysis. Second, teasing out the effects of actors’ attributes and proximity from that of existing relationships is computationally intensive. It is extremely hard to include high-ordered network structure effects in the exponential random graph model for a network of size 1840.
13. More details on the construction of the retweet network can be found in the data and methods section.
14. An ego-network consists of ego (‘focal’ node) and nodes to which the ego has a direct connection. Many existing studies have explored ego-networks but cannot untangle the formation mechanisms of follower relations.
17. The indegree of a node is the number of head endpoints of edges adjacent to that node.
18. Data collection process was constrained by the API limit, and the current study opted for a less comprehensive but feasible choice.
19. Only retweets were relevant because original tweets not retweeted by others did not contribute to the valued edges, and tweets retweeted by others were included as the retweeted of others.
20. Keywords were selected in a trial-and-error manner by examining 300 randomly-sampled tweets. They were intended to be inclusive enough to select most tweets relevant to homeowner associations and homeowner actions, whilst restrictive enough to exclude irrelevant tweets.
21. Seventy-three users did not report their locations.
22. The observed pattern might simply be the result of the research design because the influential user whom we used to define the studied follower network was located in Beijing. We could not assess which interpretation was more appropriate.
23. Exploratory analysis showed that time since account registration was insignificant once reciprocity, geographical proximity and active in tweeting were included in the model. Because the model estimation was computationally intensive, time since account registration was not included in the final model.
24. Average number of tweets per day was calculated first. It was dichotomized with the cutting point equal to the mean value. Those greater than the cutting point were coded as active in tweeting. Dichotomization had to be used to estimate the model with an average personal computer; otherwise, the computation would have been beyond the capacity of a personal computer.
25. It took about five days to obtain a final model with an average personal computer (2.3 GHz and 4G RAM).
26. Further analysis showed that tweet diffusion within a province was more likely than pure chance (p-value of QAP test < 0.001), but the correlation between the retweet network and geographical network, which was an undirected network with edges being 1 when both users were from the same province, was only 0.04.
27. The geodesic is the number of relations in the shortest possible diffusion path from one actor to another. Here, we assumed that tweet diffusion followed the most ‘efficient’ follower relations.
28. It was based on the total number of retweets contributed by users inside and outside the studied network.
29. Number of retweets, number of tweets per day and number of followers were log-transformed because their distributions were skewed.
30. The authors also used log of total number of followers as an independent variable without splitting followers into two parts. The R2 of this alternative model is 0.113, which was lower than that in model 3.
31. http://www.bbc.com/news/technology-17313793, accessed on 1 August 2012.
