Membership Turnover and Collaboration Success in Online Communities: Explaining Rises and Falls from Grace in Wikipedia

Citation: Sam Ransbotham, Gerald C. (Jerry) Kane (2011) Membership Turnover and Collaboration Success in Online Communities: Explaining Rises and Falls from Grace in Wikipedia. Management Information Systems Quarterly (RSS)
DOI (original publisher): 10.2307/23042799
Semantic Scholar (metadata): 10.2307/23042799
Sci-Hub (fulltext): 10.2307/23042799
Internet Archive Scholar (search for fulltext): Membership Turnover and Collaboration Success in Online Communities: Explaining Rises and Falls from Grace in Wikipedia
Download: http://www.jstor.org/stable/23042799
Tagged: Sociology (RSS) Online communities (RSS), collaboration (RSS), longitudinal study (RSS), membership turnover (RSS), information generation (RSS), information retention (RSS)

Summary

The authors open by describing the promise of open collaboration and social media communities to produce knowledge that might be appropriated by a brand (e.g. Threadless, Starbucks fan communities). However, the vast majority of attempts to build collaboration communities fail. These communities seem to experience high rates of membership turnover and perhaps this the high failure rates. Online communities have high rates of turnover since members are free to come and go as they please.

The predominant view on turnover in organizational science is that turnover has a negative effect on organizational performance because when people leave they take both their knowledge and their experience with the community with them. They can be costly to replace and disrupt the social fabric of the community. A second view is that the people who leave tend to be the least useful and that those who remain will benefit from their leaving. Information systems in organizations might mitigate the negative effects of turnover if the people leaving deposit all their knowledge in the knowledge-base.

On the other hand since social media platforms and wikis often record a history of every contribution and discussions in searchable archives that might serve as repositories of institutional knowledge preserved for future contributors. This might mitigate the loss of institutional knowledge form turnover. There might also be positive benefits of turnover if new members introduce new insights and knowledge. These features of online collaboration communities do not seem strong enough or universal enough to fully eliminate the negative effects of turnover. Instead the authors propose a curvilinear relationship between turnover and effective collaboration in which a moderate amount of turnover is best. Too little turnover and the community becomes rigid. Too much turnover and it becomes aimless and struggles to stay organized.

Kane et al. 2009 seems like a key citation to an earlier work by one of the authors that suggests two stages of collaboration:

1. The creation stage introduces new information
2. The retention stage preserves and refines the information through ongoing collaboration

The article finds evidence in support of this two stage model of collaboration.

They analyze the complete set of featured Wikipedia articles. A key fact about the way feature article status works is that articles may be demoted if their quality decreases like if they become out of date. This allows the authors to operationalize their two-stage model of collaboration. An article is in the creation stage before it has been featured and it is in the retention stage afterwards.

They test two hypotheses about the relationship between turnover and effective collaboration in each of the stages.

1. There will be a curvilinear relationship between turnover and knowldge creation.
2. There will be a curvilinear relationship between turnover and knowledge retention.

They use econometric methods to test these hypothesis. They use proportional hazards models predicting when an article reaches featured status (H1) and when a featured article is demoted (H2). The independent variable for both hypotheses is the average experience of editors on the article. This is the inverse of turnover in the team collaborating on the article. They also control for a number of other factors that they can measure that might be related to quality including length, the depth of the organizational structure, the ratio of external references to length, the ratio of internal references (wikilinks) to length, reading complexity, the number of multimedia elements, and the number of edits.

They also fit models using only the control variables. This allowed them to use F-tests to characterize how much turnover contributed to improving the model. They found good support for both hypotheses. The curvilinear effect during the creation stage is stronger than in the retentoin phase and improves the psuedo-R² statistic by 30%. During the retention phase the relationship is somewhat weaker, but they argue that the small effect size is still substantively important given the viewership and influence of featured articles. They standardize their independent variable when they plot the effects, this makes it easy to show that the average article is near has nearly the optimal level of turnover during the creation phase, but has less turnover than would be ideal during the retention phase. This suggests that members of the community might be more interested in generating new knowledge than in retaining and refining knowledge already created.

In their discussion they summarize their main theoretical contributions:

1. A moderate level of turnover can be best for effective collaboration.
2. Don't assume that turnover is bad
3. Collaboration has multiple phases like knowledge creation and knowledge retention.

Also provide managerial implications suggesting that communities seek to cultivate a core group of contributions while remaining open to outsiders, and find a way to archive and maintain the knowledge contributions of members that leave.

They list a few empirical limitations of their study design.

1. The study may not generalize beyond the Wikipedia setting
2. The study may not even generalize to lower quality articles on Wikipedia
3. They only looked at the creation stage for articles where the creation stage was successful (i.e. the article became featured).

Theoretical and Practical Relevance

This is a rigorous analysis of turnover in collaborative projects that looks at featured articles on Wikipedia. The study design is pretty clever and clearly demonstrates that effective collaborations have a moderate level of turnover. This contrasts with the popular view that turnover is a bad thing for peer production communities.