Tuesday, May 22, 2007

About those Youtube stats...

The TechPresident site has done an amazing job assembling the data across the internet that can be used as metrics of online support for the 2008 Presidential Candidates. They generate charts based on registered eventful.com demands, cumulative video views at Youtube, and blog posts from Technorati.com. I will be defering to TechPresident's expertise from now on when discussing the internet-based support for particular candidates.

TechPresident's cumulative video views per candidate (above) charts the actual number of individual video views for each video uploaded by the campaigns to their official YouChoose08 pages. This chart looks impressive, but is deeply flawed as a stand-alone metric for candidate support at Youtube. Below I will attempt to explain why, and will limit the discussion solely to the youtube statistics.

For starters, the video view totals include involuntary display counts achieved through embedded placements on external pages throughout the internet. Additionally, Youtube counts every time the video is played, regardless of whether one user played it a thousand times (see page-refresh schemes) or whether a thousand users viewed it only once. There is no way to measure the depth of support for a candidate based exclusively on video view counts. When these data are combined with other metrics, then a clearer picture of candidate support density can emerge.

Each candidate has one or two videos that virtually exploded in popularity, while their remaining videos received very little attention in proportion to the outliers. Remember we're trying to measure support - loyalty - to the candidates, and not the popularity of an overnight fad sensation. For that very reason I'm taking the liberty of throwing the outliers out in the text analysis that follows. Please note that the data used below is current at 5pm EDT, Monday, May 21, 2007.

Mitt Romney leads the field with 938,520 total views across 140 videos posted by the Romney campaign. But one of his videos received 440,798 views alone (46% of his total views), leaving 497,722 views across 139 videos. That averages out to just 3,581 views per video.

Hunter with 15 videos under his belt is running neck and neck with Paul as far as cumulative views go, but he too has an outlier with a disproportionate number of views (290,548 views) from his otherwise mundane 2,863 average views per video.

McCain, with 48 videos, has one video boasting 300,698 views, with the other clips averaging around 3,171 views each.

Giuliani's 16 videos show one with 49,575 views (nearly 50% of his total combined views), leaving the others to an average 3,785 views apiece.

Paul, with 17 videos in his profile, has one receiving 122,832 views, leaving the other 16 clips with an average of 15,488 views each.


Also, each candidate has a different posting style. Ron Paul posts complete videos, while Romney and Hunter post tiny sections of larger appearances that are broken down by specific topics. Let's take the May 3 debate clips for an example. Paul posted one clip, covering his participation throughout the event. This single debate clip received 122,832 views, was rated 1,255 times, and was favorited 514 times. Romney on the other hand, posted 12 individual clips, each averaging just 5,183 views, with an average of 39 ratings and 6 favoritings per clip. What can we gather from this kind of data, and how do they more accurately reflect the loyal support base of the candidates?

I spent about 8 hours combing through each of the listed candidate's videos at Youtube to pool together how many times each was rated, commented or favorited, as well as the number of subscribers each candidate has and the number of videos each candidate has posted. This information is intended the recognize the conscious action viewers took willingly to express their support, therefore more accurately representing the support candidates are receiving at the site. I hope you find the effort revealing. This is the chart that resulted from the numbers, which includes the outliers that were previously eliminated above (click for larger view):

Tossing out the outliers mentioned above, here's a chart of the averages per video, per candidate (click to view larger).

So once again while Romney has an impressive high cumulative view total, he's coming up way short on the law of averages. And once again, Paul's support seems to punch through the mediocre haze to catapult this candidate into a mainstream mainstay for the duration.

While TechPresident's data is a step up from the information we had in the 2004 presidential campaign and should be commended, it doesn't go nearly far enough, as you can see from the layers of intricate data that reflect measurements of actual free will support through voluntary behavior. I think I am not alone in expressing hope that they will incorporate a deeper analysis tool that carefully weighs and adjusts the aggregates to more accurately reflect a candidate's real underlying online support.

Update: Just stopped by the meetup.com site, and noticed of the candidate groups organized so far, Ron Paul tops the list in the general Candidate category with 2,301 members in 113 locations around the country. But, there seems to be a USA President category under which Paul is not listed. In that section, Obama has the most members at 3,210 with half the number of cities as Paul.

And over at Alexa.com, Paul's still leading the pack in website traffic, beating out even Obama.


BikeOakridge said...

Thanks for your great work!

Sonja Baumer said...

Good analysis. Indeed view count is deceptive and all kids whom I have interviewed were aware of that. "Cheat Codes" for gaming YouTube are posted everywhere online including on YouTube itself. Other techniques for gaming YouTube are "creating fictitious profiles", "deceptive tagging" etc. I think YouTube stats need to be triangulated with other data.

zchris63 said...

[GOP Candidate YouTube Stats - 06/11/07
Well now, we've got a new contender in the "GOP Frontrunner" analysis of the GOP candidate internal stats at YouTube.com:...]

anyusmoon1 said...

Can anyone (zchris63) explain what if anything happens to these results from pranksters running their finger down the spam buttons like a kid in an elevator pushing the buttons on every floor?

My comments/posts are always polite, pretty much on-topic yet I notice in google searches that some get 'marked as spam'.


zchris63 said...

[GOP Candidates YouTube Stats - 06/21/07: Here are the new YouTube internals for the frontrunner GOP candidates. It's pretty self-explanatory, but you may want to see the previous entries to see how the YouTube data is being used, and why. A discussion on the outlier data can be found here......]