Blake Narendra, a recent poster to this blog and classmate at the Fletcher School, observed that “one of the most fascinating applications for social network analysis is in the political realm.” He proposes an SNA of the visitors to the Obama White House to determine who might have the ear of the President. This analysis would certainly be interesting and add to our understanding of the inner workings of the Executive branch – but it’s import for the administration is debatable. After all, if Obama (or CoS Emmanuel) is the one lending an ear, they probably have a good sense of where they turn to most often for advice.
The purpose of this argument is not to disparage Narendra’s proposed study – in fact, I think it is a fascinating one – but rather to propose another use of SNA that is directly tied to every politician’s goal: winning an election.
Political campaigns are constantly looking for new ways to engage and persuade voters, and the Obama campaign’s use of Facebook was groundbreaking in that regard. On the day he launched his Presidential campaign on February 10th, 2007, Barack Obama’s Facebook page had only a small handful of supporters. By the time the campaign was over, he had around 3 million followers – a number that has ballooned to over 23.7 million since the election. The campaign was able to leverage Facebook and other social media tools as a way of creating a weakly-tied community around the campaign/candidate and to push messages to that community.
Study Overview
Social Network Analysis could be used to make campaigns’ use of social media even more beneficial for campaigns. Candidates are always looking to identify key influencers – those individuals that have a particular power to spread the adoption of a behavior or mindset within a network. A study of the growth of the Obama fan community could reveal some fascinating insights about individuals within the network who might be acting as key influencers. The basic idea here would be to study the diffusion of a behavior (“Liking” Obama on Facebook) within what is, admittedly, a very large network.
The eventual goals of this analysis are two-fold:
- Identify specific individuals who acted as key influencers. The campaign (or the re-election campaign) could then reach out to these individuals directly in an attempt to ensure that Obama continues to receive their support as key influencers in their communities.
- Create a predictive model based on the attributes of those key influencers that would allow future campaigns to identify key social media influencers. The campaign (or future campaigns) could then use this model to search for additional key influencers online.
Study Design and Methodology
To conduct this analysis, I propose collecting the following information – which is posted by Facebook users (to varying degrees – we would have to live with not having complete attribute information for everyone).
Network Information
- Friendship between individuals as indicated on Facebook
Temporal Information
- Knowing when individuals first “Liked” Obama is essential for mapping the diffusion of the behavior through the network
Demographics:
- Geographic Location
- Age
- Gender
- Relationship status
- “Interested In…” (a potential indicator of sexual orientation?)
- Geographical Location
- Number of Facebook friends
- Number of Facebook pages liked
- Number of Facebook groups joined
- Level of Education
- School(s) Attended – may be too varied to be useful
- Employment Status
- Employer(s) – may be too varied to be useful
- Religious views
- Political views
- Networks (as defined by Facebook)
- Favorite Music – probably too varied to be useful, but a potentially great predictor of future key influencers or supporter
- Favorite Books – probably too varied to be useful, but a potentially great predictor of future key influencers or supporter
- Favorite TV Shows – probably too varied to be useful, but a potentially great predictor of future key influencers or supporter
One of the over-riding problems with this study proposal is the fact that Facebook data is notoriously difficult to dig into (unless you are a Facebook employee). To the best of my knowledge, the analysis could not be done by someone outside the campaign, but campaign staff might be able to access enough data about individuals who “like” their page to complete the analysis.
A second potential pitfall is the sheer size of the network. 23 million fans, each with (potentially) hundreds of friends and an essentially infinite number of attributes. Mapping a network of that size would require a tremendous amount of computing power. A more practical approach might be to limit the population to a given state (or even a given Congressional district) in an attempt to reduce the size of the network to a more manageable level. The campaign could also examine data from earlier points in time, when the total number of fans was much smaller than it is today.
The third and final consideration is that “friendship” on Facebook is a relatively unsophisticated measure of connection. It does not take into account the strength of ties between individuals, which might yield a more robust analysis of how the “Liking Obama” behavior diffused through the network. However, I would be willing to bet the Facebook has – or is developing – measures of connection strength based on messages and/or wall posts between users. If the research team were able to gain access to this information, it would dramatically strengthen the quality of this analysis.
Conclusion
Overall, this may be a fairly difficult analysis to run, given the difficulties described above. We may be a few years away from developing some of the technical power needed to make this truly robust. However, the potential gains for a campaign from this type of study – even if it is not conducted in the most fool-proof and scientifically rigorous way possible – are huge. Being able to accurately identify key social media influencers could lead to a whole new tool in the persuasion toolbox for campaigns (and, of course, for other behavior-change or information-sharing organizations operating on Facebook as well).
1 comment:
Obama's FB "friends" are a tempting target, but, as you indicate, the caveats might outweigh the results. Also, what measures would you use to identify Influencers? This is essentially a contagion network, so the date of joining would be important, but is there a link between longevity and influence in a network like this? Some sort of maturity-curve analysis vs. network configuration might be interesting. That's how epidemic analysis is done (see Valente and others.). Lots of interesting possibilities, if you could get the data...
Post a Comment