The previous graph leads us to wonder whether we are viewing a power law.

A system exhibiting the power law is where power law is one where each node grows in proportion to its current size: thus, the rich get richer.  That is, if twice as many people blog about David Brooks than Bob Herbert, we'd expect Brooks's readership to grow twice as fast. And, in fact, that did happen. But it didn't happen across the board. The top 10 of 50 only had 56% of the mentions of the total list in the first period (ending 2003); this number had fallen to 47% by this year.

The mathematics behind power law often suggest a relationship where 80% of the output is produced by 20% of the inputs, and visa versa. We can't fully compute that here, because we don't really have the true picture of the population of all the pundits than a truly random selection. We'd expect to see many more pundits on the "tail."  

We should consider what no one has yet called Berra's exception to the power law. "It's so crowded nobody goes there anymore," Yogi Berra long ago said about a New York nightclub. Power-law systems are scale-free: through syndication, readership can grow indefinitely. But linking could be different, if we assume that bloggers prefer to link to less-read sources on the presumption that the may have a better chance initiating a real conversation.

Let's see how different groups did as a percentage of this punditsphere:

In the first year of TimesSelect, the percentage of references to 7 columnists out of a community of 50 of their peers decreased from 27% to 22%. Part of the slack up may well have been by the Washington Post, which increased its share. The 3:1 advantage that the Times enjoyed in 2005 was cut in half.

The percentages in this graph are based on the assumption that that our punditsphere was the end-all-and-be-all — our "punditsphere" grew by 10x between September 2005 and September 2007, while the Times share only grew by 8x (hence, the 20% loss, from 27% to 22%) . From Google's blog search, we see that the references to the word columnist grew by 24x; references to editorial, 100x, references to policy, 300x. (As noted in the beginning, this is as much to do with the growth of the blogosphere as it is to Google's ability to recognize blogs; If we use the most conservative metric of 24x, then the Times-7 lost 2/3rds of their audience, while the the Post-9 lost 3/8ths of their "audience." The link-rich didn't get richer, they got their comeuppance.

It's clear that the blogosphere simply diversified. More people blogged about local columnists, hear and abroad. And more people blogged about things other than national affairs. Still, judged against their peers, the Times columnists probably lost 20% of their blog buzz.

It may well be the Age of Diminished Expectations for Paul Krugman, and evidence that The World is Flat for Tom Friedman.

I'm afraid I haven't done enough of the statistical modeling to predict for next year. I can consider some scenarios. If the Times-7's share bounces up, does that prove that the "decline" in influence was merely temporary? If it stays flat, does that mean that the TimesSelect effect was permanent? Or is it reflective of a natural progression to the mean? If NYTimes.com has 50% more readers than the WashingtonPost.com, it would seem natural in the long run for their columnists to get 50% more attention.

One other grouping we can try (and I mean try) is to pin down columnists roughly by political tendencies:

• Liberals/Reliable Administration Critics: (Greenwald), Marshall, Huffington, Drum, Herbert, Corn, Froomkin, Dionne, Robinson, Ignatius, Kinsley, Chait, Goodman, Alter, Page.
• Moderately-Liberal-Centrists-Or-Apolitical: Hitchens, Kaus, Broder, Klein, Cohen, Cook, Zakaria, Applebaum.
• Conservatives: Malkin, Sullivan, Hewitt, Will, Kristol, Steyn, Buchanan, Goldberg, (Novak), Barone, Krauthammer, Noonan, Lowry, Barnes, Jacoby, Blankley, Gigot.

Greenwald and Novak are exceptions because of their outlier statistics. Here's the numbers we get:

The conservatives, represented by the 13 above, had great "influence" going into mid-September 2006 as compared to the prior 12 months. If that influence truly existed, it didn't mean anything in November. Starting with Senator George Allen's "macaca" moment in mid-August 2006, a series of scandals rocked the GOP, and then the Republicans lost both houses of Congress.

So it's not clear whether aggregate buzz always translates into effective influence. We don't know if the increased referencing of conservative pundits meant they were drawing more cheerleaders or critics. Entering his 5th decade at the Post, David Broder saw his blog references go up by 3.5 times in the last year. A good portion are calls for his retirement.

To come up with better conclusions, we need to start moving into the realm of solid numbers– starting with a question of how TimesSelect affected the readership numbers of the Times as a whole.