« Hypocrisy | Main | Is This the Phone I Should Buy? »

Self-Segregation and Polarization among Blog Readers

Henry, Eric Lawrence, and I have just finished a working paper that analyzes the first decent dataset of blog readers. The paper is here on SSRN or here, ungated.

The paper is motivated by normative questions about whether blogs facilitate deliberation and participation. We analyze this 2006 survey, in which about 15,000 respondents were asked whether they read blogs and which blogs they read. Some findings:

  • 34% of respondents said they read a blog. 14% of respondents named a political blog.
  • Political blog readers are, unsurprisingly, more educated, more partisan, and more interested in politics. These traits help give rise to the other findings described below.
  • Almost all political blog readers read only blogs from one side of the political spectrum. Only 6% of political blog readers named both left and right blogs. Thus, most blog readers are “carnivores” rather than “omnivores”: they like partisan red meat, as it were. This is the self-segregation that the paper discusses.
  • There is almost no overlap in the ideological orientation of readers of left- and right-wing blogs. Below is Figure 6 from the paper, mapping the ideologies of readers of some prominent blogs. The figure presents “violin plots.” The shape of the violins corresponds to where people are located on this measure of ideology. The white dots are the medians. Readers of the left-wing blogs are clustered on the lefthand side. Readers of right-wing blogs are on the opposite side. This is the polarization that the paper discusses.

big6RC.PNG

  • Blog readers are more likely to participate in politics than are people who don’t read blogs. Left-wing blog readers are more participatory than right-wing blog readers. We speculate that left-wing blogs have more fully embraced the tasks of social movements, thereby seeking to mobilize their readers.

These survey data do not allow us to make causal claims, but determining causation is is not the point of the paper. Instead, we use the observed patterns of association to draw implications for the normative value of political blogs. For most people, reading political blogs does not lead to deliberation — that is, to an exchange across partisan or ideological lines. People mainly inhabit “comforting cocoons of cognitive consonance.” But reading blogs may facilitate other normatively valuable behaviors, such as participation. Indeed, blogs like the Daily Kos explicitly want to stimulate participation more than deliberation (see Henry’s previous post).

This is a working paper, and suggestions are welcome.

(See also Henry’s post at Crooked Timber.)

Posted by John Sides on July 1, 2008 12:10 PM |

Comments

Hey, nice graphs. Why did you choose the ordering that you did? I'd think it would make sense to put the blogs in some order from left to right, but maybe there's something else you're ordering on?

Also, I'd prefer removing the bottom half from each of these "violins." I can see why you did it this way, but I find this Rohrschach effect really distracting.

Also, I don't believe some of these visual effects, e.g. the three modes for "Crooks and Liars" readers. This has got to be an artifact of how the scales are measured.

Posted by: Andrew [TypeKey Profile Page] | July 1, 2008 08:41 PM

The blogs seem to be ordered by popularity.

Two points I'm not convinced on:

What is a "political blog"? You say that they were "coded", but don't say what your criteria were. Is Monkey Cage a "political blog", for instance? I would classify it as one, unless you're restricting to partisan blogs, which would seem to miss the point. (Furthermore, the URL listed in the paper for the list of blogs does not work.)

You claim that few respondents were limited by the 64 characters provided to list blogs. I'm skeptical of this claim, but might be convinced if you gave some data on how many respondents approached the limit.

Posted by: Dylan Thurston | July 1, 2008 09:54 PM

Dylan: The Monkey Cage would be a political blog, by our definition. (The URL is now fixed. Apologies.)

Among those that actually listed a blog or blogs, only 7% used the entire 64-character limit. The average was about 23 characters.

Posted by: John Sides | July 2, 2008 11:54 AM

It seems to be a bit of a self-fulfilling prophecy. Here, blogs general tendency is to draw like for like...people taking the time to read (and participate in) blogs that are of interest to them. In politics, it is not surprising that the trend is to find out first what is going on in "your" party or with "your" candidate before checking out the other side. With the number of blogs out there, and the tendency (expectation even) to link references in one blog to another (or to news articles, videos, etc), readers must prioritize time and content. So right-leaning readers find conservative leaning blogs and vice versa, trending the content of those blogs further along the political axis one direction or another simply through their participation.

I also would be interested in the addition of some analysis of the tendency of political blogs to sit to the more extreme ends of the spectrum in order to function. Bipartisanship is all well and good, but a a general knowledge of social dynamics of the internet will force the recognition that if politics (and religion) are contentious topics in polite conversation, on the internet they are potential flame wars. I can't imagine a moderate "blog" even surviving a presidential election cycle, as the conversations degenerated into more polarized views and personal attacks.

Godwin's Law may be almost 20 years old but it's still valid. So, there could be an argument that the nature of political blogs to polarize themselves is also a matter of maintaining useful and applicable conversations (albeit biased) to insure blog/site integrity.

Posted by: Heather | July 2, 2008 12:02 PM

Andrew: Dylan is right. The blogs are ordered by popularity, where popularity is defined by the number of respondents in our sample who report reading the blog (see Figure 2 in the paper). The most popular blogs are on the bottom, as in Figure 2. We could flip it and put most popular at the top, but at a minimum, we'll clarify it in the figure note.

We like the violins, but we'll revisit the choice.

The multimodal appearance you note stems from the fact that the scale we have has eleven values, so the distribution tends to be lumpy in homogeneous groups like self-identified Crooks and Liars readers.

Thanks.

Posted by: Eric L | July 2, 2008 12:28 PM

Eric,

If you have discrete categories I think it would be clearer as histograms. The bumps must represent some weird artifact of post-processing the discrete data to look continuous.

Posted by: Andrew [TypeKey Profile Page] | July 5, 2008 09:37 PM