shouting loudly

[edited to remove a few sloppy suppositions.]

Blog research in political science is still pretty much in its infant stages.  There are maybe one or two dozen people who work on these topics, and most of us are (like myself) very early in our careers.  The lack of a vibrant research community means there aren’t institutional spaces where we can hash out deep theoretical differences.  So, I guess I’ll go with the next-best-thing and publish my complaints on this low-traffic blog.

I’m currently reading Richard Davis’s new book, Typing Politics: The Role of Blogs in American Politics (Oxford University Press, 2009).  This is Davis’s fifth book, all with major university presses, making him arguably one of the deans of this nascent research community.  In it he repeats a couple of major flaws endemic to political science research on the blogosphere — by which I mean they are mistakes that shouldn’t be made, and that no one else seems to make.  Consider this post my first attempt at calling these flaws out.  The three issues, as I see them, are (1) blog structure, (2) left-right balance, and (3) how we conceptualize “elites.”

Blog Structure

Davis claims to offer a detailed analysis of seven “influential” or “A-list” political blogs: DailyKos, Eschaton, Crooks and Liars, Instapundit, Little Green Footballs, Wonkette, and Michelle Malkin.  He also discusses Huffington Post, Talking Points Memo, and a few other major blogs.  Davis attempts to describe the blogosphere across two dimensions — size (influential vs common) and ideology (left vs right).  Therein lies, for me at least, a major problem that is simply inexcusable in 2009.

Here’s a good way of phrasing the question.  As an empirical matter, how many blog posts appeared on DailyKos in September 2006?  Davis includes this as a datapoint in his book, so it’s plenty relevant.  He thinks there were 429 posts — fewer than Instapundit, Eschaton, and Crooks and Liars.  I compiled data on the lifetime posting rates on DailyKos for my MPSA paper, “stability and change in the poltiical blogosphere.”  Downloading the entire dataset of dKos blogposts, I find there was somewhere between 8,000 and 9,000 posts that month (i’m eyeballing the graph, because this is a spur-of-the-moment blogpost.  I could get the exact number in less than an hour, though).  That’s a pretty sizeable difference, to say the least.

The difference is that I’m counting user-diaries, which DailyKos allows and the other three sites don’t.  Hell, Instapundit doesn’t even allow comments.  Those diaries are blog content.  Diaries that reach the recommended list receive hundreds and sometimes thousands of comments — more than front page posts do — and that’s a strong indicator that they play a central role in the site.  I argue in Understanding Blogspace that this difference in site architecture is effectively a difference-in-kind, that community blogs act as quasi-interest groups and forums for collective action.  If we group DailyKos with Instapundit, we ignore this distinction entirely.  If we lop off 97.5% of the content produced on a site (and do so while providing no indication or rationale to the reader), then we get a pretty badly skewed picture of what’s actually happening in the blogosphere.

Furthermore, this serves as an important indicator that the “elite vs common” distinction just isn’t good enough.  Instapundit simply does not have more posts/month than dKos.  That is factually inaccurate.  DailyKos has more posts/month than the top 25 conservative or progressive blogs combined.  That makes it a hub in a power law distribution, which is network-theory-speak for “way, way bigger than any of its competitors, and increasingly so.”  And difference of degree ends up equaling a difference-in-kind.  The size differences within the “A-list” are at least as noteable as the size differences between A-list and “common.”

DailyKos community members raise millions for the slate of candidates they endorse, often contributing more to their top candidates than the DCCC does.  There is no conservative equivalent.  That matters profoundly for the skeptical analysis of blog influence that follows in Davis’s book.  He treats DailyKos as if it = Markos Moulitsas, rather than Markos + 12 frontpage editors + thousands of active diarists.  If we limited the Sierra Club to Carl Pope, we would similarly find that his influence is pretty seriously curtailed, but then again no serious press would print a book that made such a basic flaw.  Yet Davis is just continuing an error-prone tradition that has plagued blog research since our earliest days.  Blogs in 2000 were 1 blog per blogger.  Beginning in 2003, we had community blogs.  Since then we’ve also gotten institutional blogs, and further divergences.  Those differences should be the subject of scholarly attention, even if they make our research designs a lot more complicated.

Left-Right Balance

Davis reiterates Adamic and Glance’s semi-classic paper from 2005, which conducted a large-scale hyperlink analysis and found that the left and right blogospheres inhabit separate and rarely-overlapping neighborhoods.  Yes, this is correct.  But he then goes on to walk a thin narrative tightrope, acting as though the left and right blogosphere are essentially equivalent to one another.  For a book which claims to provide solid empirical data, this is extraordinarily difficult to justify.

After the 2004 election, there were a spate of research papers that tried to measure the blogosphere, mostly using hyperlinks, and found that the left and right were either equal or that the right was slightly stronger.  This latter point came from Robert Ackland, based on the conservative habit of including each other more frequently in blogrolls (why this supposedly operationalizes power is beyond me).  You could maybe make the case that this parity was real in 2004 — I don’t have the data, so I can’t refute it.  But in 2008?  That is simply ridiculous.

In one memorably passage of the book, Davis writes, “[leftwing blogs] helped raise an estimated $2.3 million for Democratic candidates in 2006.  Similarly, conservative blogs promote donations through Slate Card, the Republican equivalent of ActBlue.”  Notice how he left out a dollar figure for SlateCard?  That’s because the site never really took off.  It has raised about $650,000 total, as opposed to the $93 million raised through ActBlue.  What’s more, Slatecard is one of several conservative attempts to replicate ActBlue, all of which have essentially failed.  The only successful conservative “moneybomb” came from the Ron Paul crowd, and they were actively ridiculed by the network of elite conservative bloggers.  These differences are empirically demonstrable through the Blogosphere Authority Index (which was the whole reason I created the thing).  Again, my MPSA paper and my original BAI publication specifically demonstrate that the elite progressive blogosphere is far larger than its conservative counterpart, that this gap is growing, and that conservative bloggers have had a particularly tough time replicating progressive successes in online infrastructure.  Over $93 million has been raised for Democratic candidates through ActBlue, and the major beneficiaries have been those candidates who received support from the major progressive blogs.  For a book on the role of blogs in American Politics to  ignore these points, papering over them continuously through selective use of the available data, is, to me, deeply problematic.  These points were non-obvious in 2004.  They are exceedingly obvious in 2009, and I would suggest that if a scholar wants to make the case that there is little difference between the two blogospheres, the onus is on him/her to explain away these empirical facts.

Davis also mistakenly attributes a few conservative talking points as facts in his description of individual bloggers, claiming that Moulitsas endorses candidates based on whether or not they give consulting contracts to Jerome Armstrong (that has been disproven to the point of ridicule) and that Michelle Malkin had to move homes after her angry critics posted her address online (this is true, but it occurred as a response to her doing the same thing to a set of anti-war protesters at UC-Santa Cruz).  Granting Davis a wide berth, I would say that the effort to represent both networks of bloggers as equals leads to a skew in favor of the conservatives.  More broadly in the literature, we as a community have failed to capture this conservative deficit because the data is legitimately pretty daunting.  I designed the BAI specifically to help scholars get around this problem, and I am hopeful that it will lead to greater accuracy in the literature over time.

Conceptualizing Elites

Davis spends fully three pages (pps 40-42) describing the results of Matthew Hindman’s “blogger census” conducted in 2004.  Matt contacting the primary blogger from 75 elite sites to find out their demographic characteristics, and concluded that the elite blogosphere replicates the demographic disparities found elsewhere in society.  This is the core of his argument that the blogosphere is “just different elites,” to which I have often replied “yes, but they’re different elites.”  Matt describes the survey himself in The Myth of Digital Democracy (Princeton, 2008) on pages 118-128.  A little something is lost in replication, as Davis does not include all the methodological explanation that Hindman does, thus enshrining the 2004 census as deeper fact than it actually is.

My problem with this census is twofold.  First, there’s the issue of how it handles mega-hubs like DailyKos.  Hindman decides in cases of multi-author blogs to take the individual with the highest number of posts.  I certainly understand why, in 2004, that would seem a reasonable design choice.  But nonetheless, it leads to the conclusion that the blogosphere is, for instance, overwhelmingly male dominated.  Well is it?  DailyKos has 20 frontpage editors (some are no longer active).  All of them have equal access to the 850,000 visitors the site receives per day.  13 are male, 7 are female.  Instapundit has 1 editor, and he’s male.  So should that be scored as 2 male, 0 female, or 14-7?  Adding the editors or lead contributors of other sites similarly changes the racial complexion of the elite blogosphere, though it does remain majority-white, to be certain.

The blogosphere remains largely well-educated and otherwise demographically advangated (high education and white-collar job status are correlated with various attributes that make sophisticated blogging easier).  But to say that it is “just elites” strikes me as wrongheaded specifically because there is mobility within top community blogs like dailykos (and huffington post, MyDD, FireDogLake, and OpenLeft).  Steve Singiser just joined the DailyKos editors list last week.  he was a members of the community for years, contributing diaries as a volunteer in his spare time.  His diaries were popular and highly recommended, and this led him to eventually be elevated to front page status, with a platform to the gargantuan reader-base.  That’s meritocracy.  Not a perfect one, and not an egalitarian one (as far as I can tell, he’s another well-educated white guy, albeit a high school social studies teacher), but far more fluid and meritocratic than other elite systems of influence.

This is a challenge for blog researchers going forward — one that I don’t fault anyone for not getting right just yet.  There is significant mobility in Dailykos, and none in Instapundit.  There is some in other sites.  It depends on blog structure and community norms.  That is tremendously interesting variance, laden with questions that we should be asking.  I hope that Hindman’s early “census” doesn’t become so foundational to the literature that we fail to ask such questions.  It was a fine study for 2004, a little dated by 2008, but relying uncritically on it paints an increasingly inaccurate picture of the blogosphere today.  And what’s more, it is a picture that political scientists get wrong, while everyone else gets it right.  Computer scientists know about diary structures, etc at this point.  They ask questions about community scaling and related code-based algorithms.  For our research to remain actively relevant to the impact of blogging (or other online political action) on politics, we need to keep a close eye on how the hub sites themselves have grown and changed.

That’s all I have to say at the moment (probably far more than I should).  Suffice it to say, I didn’t love this book, and I think it repeats some mistakes that need correcting.  I’m hopeful that as the scholarly community grows and builds institutions for airing these theoretical grievances, we’ll start getting more of these basic theoretical points right.