Reading the Web Political Tea Leaves

About a month after playing around with Google Trends for Web Sites I noticed that conventional wisidom and actual site traffic didn’t always match up.  For example, main stream media news sites have less traffic in many cases than popular blogs (check abcnews.com or cbsnews.com) against some popular political blog sites.

This got me interested in seeing what was going on in the traffic of the presidential candidate sites.  Does Obama rule the Internet?  Well try it yourself and see (http://trends.google.com/websites?q=barackobama.com%2C+johnmccain.com&geo=all&date=all&sort=0) – the conventional wisdom of the Obama campaign “getting” the Web is clearly upheld.

However, as you drill in you start to see interesting peaks and pinches.

This got me wondering if there is some value to looking at Web trends to determine what people are thinking about the election particularly as related to various events in the “real world”?  My answer was clearly likely yes – it is just a communication medium.  However, people still hold the Internet and Web in a special light so it seemed a ripe topic for investigation.

Looking at Hitwise the same types of trends were seen with Obama far eclipsing McCain.  Also similiar peaks and valleys were seen over similiar date ranges.

Regardless of what site I visited the general trend lines were the same.

After sharing this information with a variety of folks it was clear an article was in order.  Given the nature of the domain I enlisted a recovering Poly Sci wonk who happens to work with me and we went about looking all over the Web for what it could say about the election by reading network data.  What we found is briefly summarized here

• Obama is more popular online than McCain by a significant margin
• Web 2.0 environments heavily favor Obama to the point of showing major sampling bias
• The trend lines match the offline polls but without the wide margins

It isn’t partisian it is observation but already the fur is flying. The full piece can be read at Network World ( ). It would seem for the comments getting to us that the Internet itself is a sampling bias.  That seems a bogus argument today, but ok.  I’ll counter with the phone as a technology sampling bias in two ways, non-phone users and cell phone users.  While the first group might be lesser the later clearly age biases polls to those who retain land lines.

The next argument against the piece seems to revolve around the idea I have called the visitation fallacy for the last view years.  The protesting readers assume larger traffic could well be bad traffic or people vetting the unproven candidate?  We agree, intention isn’t clear from simple visit or view.  A lot of page views might be a happy user or a sad user.  However, there were some problems with the thinking.  The trend of polls and happenings in the “offline” world seem to map too clearly.  For example, at the height of the RNC convention the attention was on McCain and the height of the DNC event on Obama.  When the economy melted down the attention went to Obama following   I agree, lots of visits/page views for a site doesn’t equal happy or supporting users.  We noted that in particular the blog posting measurements showed more equality as there was pro/con articles from both sides.