A Purposive Sampling Method for Internet-based Research

Yesterday, I wrote a post about Gizmodo readers commenting on Transformers costumes. I linked to Joachim Bengtsson, who commented to ask why people feel the need to qualify their praise of the elaborate costumes by declaring that the costume designers were really nerdy (not to mention that the video had been posted in Gizmodo’s “Too Much Free Time” category). And I wondered aloud: If there are tens of millions of self-identified geeks in the US alone, and if (it seems pretty safe to assume that) most of them aren’t doing cosplay, fan fiction, machinima, or the other super-involved activities that receive a disproportionate amount of attention among fan scholars but get derided as “too nerdy” by so many fans, then what are these geeks doing that’s so geeky? Well, I think I stumbled upon one way to find out.

For my original post, I wanted to give a link back directly to the fellow who left a thoughtful comment on Gizmodo, so I went back to the post, clicked on his name, and found his home page. And then I had the answer to the above question: right here, I found out what one geek has been spending his time on, including five blogs and a number of software projects. It’s not cosplay or fanfic, and it’s not even necessarily related to media fandom per se, but it’s certainly some interesting creative output. I emailed Joachim to ask if he wouldn’t mind being used as an exemplar in this post, and he happily agreed.

Perhaps this has been done before and I have missed it, but this seems like a potentially useful sampling technique for internet-based research. Most internet-based ethnography I have seen has drawn from forums, newsgroups, or online games, though blogs offer a different sort of environment. Online games definitely worth studying in their own right, but if you’re not specifically studying gaming audiences, you’re best served elsewhere. Forums tend to need to be very narrow in order to keep conversation on topic and draw any audience at all—so again, you’re okay if you can find a forum that seems along the lines of what you’re looking for, but even then you might only be able to find a small group of people actively posting. Blogs draw very large crowds, however, and many of those readers post comments as well.

So: go to a blog that allows registered commenters to link back to their home page. Follow those links back to people’s home pages. Suddenly you get a sample of personal blogs, pages listing people’s pet projects, websites for startup businesses, and so on. For research that need a more representative sample of internet users, this wouldn’t be appropriate because of the bottlenecks involved: how many blog readers actually comment, and how many of those link back to a page, and how many of those put anything useful up on their page. Still, I imagine that what you do find could still be useful, and knowing what percentage of commenters link back to something is potentially interesting information in itself. For research like mine, it’s an easy way to get at what particularly active blog readers’ interests are, and how these interest communities overlap. Plus, it’s relatively non-intrusive, so long as you leave everyone anonymous in your write-up.

My deep-seated academic urges tempt me to come up with a clever term for this sampling method, but I’ll sit on that because (a) I don’t know if this is useful to anyone besides me, and (b) someone else might already be doing this for all I know. Whatever the case, please feel free to let me know any thoughts/criticisms, and thanks again to Joachim for agreeing to be my exemplar. Also, check out the interesting follow-up at his own blog.

3 thoughts on “A Purposive Sampling Method for Internet-based Research

  1. Sounds like ‘snowball sampling’ to me – using one subject to refer others to your project, although as the researcher you are the one tracing contact links, not your initial subject. Usually used with hard to access populations where the potential access to subjects outweighs the non-representativeness of your sample, which seems to fit your research fairly well. Although you probably run into a self-selection issue with who actually registers and comments on the originating blog… maybe take a sample of blogs in your interest area as well, to spread out the impact of the primary blog sources?

  2. Yeah, I wasn’t sure if snowball sampling necessarily had to use people’s social networks or if other networks of information still counted. I’m also wary about using the term “snowball sampling” to refer to any method of finding information that I don’t have another name for, especially since the process described above seems (to me) much more planned out and purposive than the “dumb luck” approach I was writing about some days ago.

    Whatever the case, I’m starting to suspect that the above technique is much more limiting than I originally expected, as very few commenters on Gizmodo seem to be linking back to their own pages. It is interesting, though, that you can see an entire history of one person’s comments, which is another way of constructing a sort of profile for a sample.

  3. Hey! I think, in all probablility, very close to hit. Other researchers on the internet have classified the studies as Non-Random (Non-Probabilistic) Sampling and the studies were using Convenience Sampling just the way reserachers might go in a room and ask for volunteers.

    Of course, Puposive Sampling is under the same heading of Non-Random sampling. The type of sample is in the same gendre and I agree with your thinking, the Internet can offer a very specific group of participants for online research.

    Unfortunately, I can not see that is snow ball sampling, but it could be used on the internet if one referred you to the next. It is also a non-random method.

Comments are closed.