In an earlier post on this topic, I looked at 91 sites on my blogroll that had open Sitemeter statistics. I found that an an inbound link is correlated with 11 additional visits per day. One might believe that links cause visits, or, as Professor Volokh noted in an email quoted in the linked post, that visits cause links or that some third factor causes both links and visits. (New Poll Shows Correlation Is Causation.) Data are not currently available to test which of these hypotheses are correct, though I'm hoping they will be soon.
Thanks to the amazing N.Z. Bear, I now have both traffic statistics and the number of inbound links for 682 weblogs (excluding identifiable duplicates). Of those 682 sites, I've been able to identify approximate first-post dates for 651. Some of these are likely to be inaccurate for blogs that have moved and neither imported their earlier posts nor linked to their earlier incarnation(s). (Dean Esmay's Blogspot Jihad was a boon to readers, but made this research somewhat more difficult.) I've also identified which of the 682 sites are hosted on the much-lamented BlogSpot servers.
In a comment on the post linked above, CGHill of Dustbury fame hypothesized an inertial effect, i.e., that visits are correlated with blog-age. As a first result, I looked at the correlation between blogs' age and traffic. On average, each month a site has been in existence is correlated with a (statistically significant) additional 26 daily average visits. So there's some preliminary support for this theory.
Next, I looked again at the correlation between inbound links and traffic. I find that each inbound link is correlated with a very statistically significant additional 12 visits per day. (This is similar to the 11 visits per day I found in the earlier analysis. The increase may be explained by the increased popularity of weblogs over the last couple of months.) If it can be shown that links cause visits, the number of inbound links alone explains 60% of the variation among blogs in the number of visits.
Adding blog-age to the analysis, I find that each inbound link is is again correlated with 12 additional visits per day and each month in existence is correlated with a statistically significant 15 fewer visits per day. Negative inertia?
Finally, I looked at the relationship between daily traffic, inbound links and whether or not a site was hosted on BlogSpot. Again, each inbound link was correlated with 12 additional daily visits. Sites on BlogSpot received a statistically significant 171 fewer visits per day than those hosted elsewhere. However, it's fairly common for sites to mature and be moved off BlogSpot (the "training wheels" phenomenon). Adding blog-age to the analysis, I find that, once again, inbound links are correlated with 12 additional visits per day, each month in existence is correlated with 14 fewer visits per day, and sites on BlogSpot receive a statistically insignificant 60 fewer visits per day.
The sites with the highest traffic given their age and number of inbound links ("readers' bloggers") are Gizmodo, Instapundit, Tom Tomorrow, Eschaton (Atrios) and Daily Kos. On the other hand, the "bloggers' bloggers'" are The Truth Laid Bear, A Small Victory, IMAO (I expect this might speed-up the delinking process.), VodkaPundit and Outside the Beltway.
In this post, OTB's James Joyner express surprise that his traffic appeared uncorrelated with the number of inbound links to his site. If links cause visits, my overdue explanation is that, at present, he's a "bloggers' blogger."
For what it's worth, my site is also very much of a bloggers' blog, receiving a rather-astounding 600 fewer daily visits than would be predicted. Thanks for all of the links, folks. Hopefully, the readership will follow...
I was going to post a few mildly interesting (to me, at least) charts of some of the underlying data, but this post is already too long. Perhaps I'll put them up in the days to come.
Your feedback is eagerly sought.
Posted by oscarjr at July 14, 2003 10:28 PM | TrackBackI believe I have found a profound error in Truth Laid Bear's site statistics ranking which may render your data questionable. I'm going to write something up about it tonight.
Dean
Posted by: Dean Esmay at 11:52 PMAgain, intriguing stuff. I'll be interested to see what Dean's post is about.
I have found that my traffic has been lower lately, corresponding roughly with the start of summer. Not sure how much of a causative factor that is. I seem to be down to a certain group of primary readers who visit no matter what, and some who used to read me don't seem to any more.
I've also noticed that my trafic is driven by posting. I post, there's a surge from people who have watched for recent updates on blogrolls. After a couple hours, it dies back to minimal traffic. Posting nothing in a day puts me below 100 hits. Posting prodigiously still sometimes doesn't get me over 150. The time of day at which I post matters. I've developed a habit of posting heavily between midnight and 4 AM. That loses most of the "recently updated, let's go see!" traffic.
I've been getting fewer meaningful links than I was for a while, but I can understand that in light of my posting having been spottier and lower in quality for a number of weeks. I seem to be back in better form, though I've been doing a lot of quick link posts. Those might get who I'm linking some traffic, but they don't generate mentions for me the way original commentary does.
Anyway, just random thoughts that are neither here nor there. I look at my traffic flow regularly and tend to think a lot about cause and effect.
Posted by: Jay Solo at 12:10 AMDean:
I look forward to your post. If the alleged error is in NZB's traffic statistics, I doubt that it will change my analysis substantially, as it mostly validated the numbers I ran previously.
If the alleged error is in his inbound link numbers, it might be more important. That said, the numbers seem to correlate well with Technorati's data, so I'll be eager to see.
Jay:
I think you're absolutely correct that posting frequency, timing, and quality likely matter greatly. If I, or Mr. Bear, could readily measure such things, I think the explanatory power of the analysis would increase substantially. The data are difficult to measure, of course, though I seek to do so.
Thanks to both of you for the input.
OJr.
Posted by: Oscar Jr. at 01:48 AMOh: One amusing thing is the TTLB blog itself. It is ranked in the top 10 now in terms of links, but the URL is for the Ecosystem rather than the blog portion of the site. The Ecosytem is generating huge buzz, because bloggers love to navel gaze. The ultimate blogger's blog, I guess.
Posted by: James Joyner at 09:12 AMActually, I think the URL is for the main site, but I'd imagine NZB counts links (and visits) to both. Still, it's a great way to draw both links and traffic. That Bear's no fool.
Posted by: Oscar Jr. at 09:24 PMInteresting analysis. One suggestion: every link is not the same. Links in posts drive much more traffic than blogroll links. Links in posts from high-traffic sites drives lots of traffic to their links. Blogroll links on lightly-visited sites drive very little. TTLB's stats don't really allow you to check the difference EXCEPT that we can assume links to particular archives are post links. It would be fun to check out for someone who has the time.
Posted by: Allen Brill at 07:16 AM