October 27, 2008, 8:31 PM — When you take a close look at the traffic patterns within the Web 2.0-based community, the popularity gap between the two presidential candidates increases. Obama's favored by a 4-to-1 margin compared with the 2-to-1 margin when we looked at other Internet Web traffic trends.
However, we can likely attribute this to a clear case of demographic sampling bias, probably based in part on the age cohort for such sites skewing younger than the broader Internet audience, and thus being more likely (if the demographic breakdowns in traditional, offline polls are to be believed) to favor Obama over McCain. For example, looking at MySpace and Facebook "friends," the advantage noted for Obama in the site traffic numbers widens considerably, with about four times as many users "friending" Obama as McCain.
A particularly vivid example of the possible sampling biases in Web 2.0 type metrics is provided by techPresident's chart of candidates' supporters who are using the online social networking portal Meetup.com. Here, not only do Obama supporters outnumber McCain supporters to an even greater degree than we saw with Facebook and MySpace, but even more notably, the supporters of third-party candidate Bob Barr actually eclipse McCain supporters and rank a close second to Obama supporters. Clearly, this particular Web 2.0 metric is measuring something quite different than what we find in traditional polling results.
Clearly unrepresentative anomalies like this one aside, the techPresident site is a particularly rich source of Web 2.0 metrics of candidate popularity. Its partnership with online video analytics firm TubeMogul, for instance, gives techPresident the ability to show detailed dynamic visualizations of total video views for each candidate on YouTube, as in the following pair of charts:
The day-by-day data here is somewhat "noisier" than some of the other metrics we have looked at, with wide swings between the two candidates' video viewership numbers (compare it for instance to Google's daily traffic numbers for the candidates' Web sites). Intuitively, YouTube viewership seems more responsive to day-to-day events (in both traditional and online media), resulting in the greater variability. Still, the overall trend is clear, as the cumulative viewership trend lines demonstrate, with Obama building up a long-term advantage that closely mirrors his 4-to-1 advantage in the social networking sites.
On LinkedIn we see a markedly different pattern. While both candidates have the maximum viewable 500+ friends, note that the McCain campaign has recommendations and has made one (Sarah Palin) while the Obama campaign has none. (Read a slideshow of 12 tips for safe social networking.)
Readers might suspect the business focus of LinkedIn is at work here, but in our view the numbers are too low to make that conclusion reasonable. Given these remarkably low totals -- including zero recommendations for Obama -- we suspect campaign editorial decisions may be minimizing the use of this particular network for political expression. This is likely analogous to McCain's near-invisibility in other Web 2.0 systems notably online event systems like Meetup and Eventful.
In general, our research into Web 2.0 type metrics forces us to conclude that, when using the Internet for social-observational purposes as we are doing here, following the law of large numbers seems to be a good way to control sampling bias, whether it results from the demographics of self-selection or conscious decisions on the part of the campaigns about where to focus their efforts.
"Can the Web predict the next president?", Network World (US), Oct. 27, 2008.
"Election '08: What's in a domain name", Network World (US), Oct. 27, 2008.
"Palin piques the blogosphere", Network World (US), Oct. 27, 2008.
"Hot search terms: Joe the plumber, 'lipness test'", Network World (US), Oct. 27, 2008.