Cognitive Daily

Today’s analysis of the Blogger SAT Challenge results is the one I’ve been looking forward to the most. After subjecting 109 people to a sample question from the SAT writing test, we’ve learned that bloggers are dumber than high school kids (though there’s some reason to question that analysis). Our participants, most of them bloggers, didn’t fare nearly as well as high schoolers.

But bloggers have all sorts of excuses to explain their poor results: They were multitasking at the time; they hadn’t spent 18 months in an SAT prep course like the high schoolers; the judges don’t “get” sarcasm. Fine. Let’s compare the bloggers to another group with all the same disadvantages: the non-bloggers who participated in the challenge. Of the 109 participants, only 63 were genuine bloggers with listings in Technorati. So, how did they compare to the non-bloggers in the group?

Actually, not too badly:

The bloggers’ average grade on the test was a 3.13, while non-bloggers averaged just 2.60. Bloggers still didn’t score better than high schoolers: Their 5.74 average SAT-format grade was much lower than high schoolers’ 7.2 average. Below the fold, I’ll present a graph of three different comparisons between bloggers and non-bloggers.

i-27f54737d2c6579885e3f6a9ff51d22f-blogger1.gif

As you can see, bloggers did better in all three measures (though only the “average most popular rating” difference was actually statistically significant). But this brings up another question. Can we find scoring differences among the bloggers as well?

I looked up the Technorati rankings for each of the 63 participating blogs. I also recorded the number of blogs linking to each of these blogs. Here’s a chart showing the relationship between the number of links to each participating blog and that blogger’s score on the Challenge.

i-cd0e75fd6047570be3bf0ab00b9d8013-blogger2.gif

The trendline shows that scores do appear to be higher when blogs have more links to them. Since the number of blogs linking to a particular blog is an accepted measure of a blog’s popularity, we can say that more popular bloggers tend to score higher on the SAT test. Indeed, the number of links to a blog is positively correlated to grade with a correlation coefficient of .36 — a significant correlation!

The chart of average reader ratings tells a similar story:

i-214d35155347e7c17d720f26f735f59a-blogger3.gif

Once again, there’s a significant positive correlation (r=.32) between the number of links to a blog and the quality of the blog author’s SAT challenge essay.

What this data doesn’t tell us is whether blogging makes you a better writer. It might be that better writers choose to write blogs by virtue of their writing ability. Nonetheless, these results may be somewhat of a vindication for the bloggers. They may not be as good as high schoolers, but at least they’re better than all those other schmucks out there!

Comments

  1. #1 Janne
    October 6, 2006

    How about correlation to time since they started blogging?

    Blogging time should correlate with writing ability, simply because practice makes perfect. But it should also correlate with popularity, since it takes time to build an audience, and since becoming a better writer over time should also mean more enjoyment forthe reader.

    By that token, a “naive” blogger, the non-bloggers, could perhaps fit into that data as the close-to-zero writing experience group?

  2. #2 ck
    October 6, 2006

    It would also be interesting to look at frequency of posts (along with the “practice makes perfect” theme) and the content of posts. Personal ramblings on MySpace would correlate with a different score, I think, than an academic blog.

  3. #3 pete
    October 6, 2006

    couldnt get my brain going to even think about the graphs, how about good bloggs are good to read if you like the subject they are talking about and blogs that are boring your not intrested in because the subject isnt important to you

  4. #4 Dave Munger
    October 6, 2006

    ck and Janne: I think both frequency of posting and length of blogging career are probably related quite reliably to blog popularity, so we may have those issues covered.

    ck: Hmmm, yes, a content-based analysis might be interesting, but I think I’ve about hit my limit on this subject.

    Pete: Yes, clearly writing isn’t among your interests. Perhaps you could take up fishing.

  5. #5 Pete Mandik
    October 6, 2006

    I don’t know whether I am a good writer. I do know that I’m a better writer than I use to be. I also know that the main thing that improved my writing was getting into email flame-wars with my fellow grad students back in the day. There’s nothing like having your ego converted into text and put on the chopping block for immediate feedback in front of an audience to really condition you into being (more) worth reading (than you were before).

    Why shouldn’t the same be true of blogs?

  6. #6 Sunil Bajpai
    October 7, 2006

    Could it be that beyond a point, better writing (as judged by an expert’s score) turns out to be negatively co-related with a blog’s popularity? Perhaps, because it’s too difficult or too different from what their audience can relate to?

  7. #7 lylebot
    October 9, 2006

    The second figure is not really convincing. What happens to the trendline and correlation if you remove the point with a grade of 6?

  8. #8 Dave Munger
    October 11, 2006

    Lylebot, the correlation diminishes substantially (to r=0.21) if you remove the outlier, but in principle, I find little justification for doing so. The same essay was rated highly by both the expert graders and by the blog readers. *Someone* has to score the highest, and if that person also has the most popular blog, doesn’t that say something about the relationship between blog popularity and writing ability?

    The big problem with this, of course, is that the essay in question was written by me.

    That said, I think a case could also be made for removing a second outlier, the essay from the fourth-most-popular blog which received a score of 1. If you remove both my essay and that one from the analysis, the correlation moves back up to a fairly respectable level (r=0.28), still a significant correlation.