Constructing a purposive interview sample of bloggers

David Brake

Constructing a purposive interview sample of bloggers

Internet Research 8.0, Vancouver October 18-20, 2007

By David Brake

London School of Economics

http://davidbrake.org/

David Brake

Why interview personal webloggers?

• They write what appear to be ‘private’ statements in a ‘public’ space

• This study seeks to understand the practice through the self-reflexive accounts of practitioners (not always provided in blogs themselves)

David Brake

Why do so face to face?

• Primarily to get richer more lengthy accounts

• Lüders who mixed instant message and face to face interviews found the length of her face to face interview transcripts was three times that of the IM transcripts

• Also provided verification of age and gender

David Brake

Sampling problems of previous studies

• “All informants were well-educated, middle-class people either in school or employed in knowledge work or artistic pursuits.” (Nardi et al 2004)

• “The vast majority … identify as middle class. Nearly all of them are university educated” (Reed 2005)

• “The interview participants were almost entirely ‘information professionals’ in some form.” (Brady 2006)

• “generally well-educated professionals and business owners; people involved in higher education or actively pursuing new educational opportunities.” (Lenhart 2006)

David Brake

Reflective of blogging population?

• Using Dec 2005 Pew data:• 8% of US Internet users overall• College educated - 6.1% blog• Lowest of categories - less than high school =

13.8%• Household income $50-75k = 5.2% blog• < $10k - 18% blog• Age - 24.7% of 18-24 blog but 5.5% of 55-64

y/olds blog

David Brake

Representative sample?

• Characteristics of UK webloggers then unknown

• Numbers to be interviewed not sufficient to represent full spread

David Brake

The solution?• Used purposive sampling (Chadwick et al.

1984 pp 65-66) - that is:

• Sample constructed to maximise variety - in this case of age, social class, education, blog audience level and attitude to readers

• Wish I had heard Dan Li and Gina Walejko’s presentation here two days ago two years ago!

David Brake

Four step approachI Use search engines to find

contactable personal webloggers

II Select webloggers for initial contact from search results

III Send questionnaire and select interviewees based on questionnaire results

IV Interview respondents f2f

David Brake

Step 1

• Search terms constructed using Google to find London-based blogs hosted by blogger or LJ with email addresses (using search of profile pages)

• Eg site:livejournal.com "User Information" London e-mail -Ontario

• 510 sites examined (starting at bottom of search results to avoid popularity bias)

David Brake

Step 2

237 sites were selected matching criteria:• in English• by one author• classified as ‘personal’ (based on reading first

page)• updated in the last month and contained at least

five entries• author appeared to be in or near London• an email address was available.

David Brake

Step 3

• Sites from Step 2 emailed an invitation to do a short (21 question) web questionnaire

• 150 responses to online questionnaire, from which 75 were invited to interview (invitations sent until desired breadth of response achieved).

• Note: questionnaire results were therefore not anonymous - at least not to me. Questionpro enabled tracking and reminder emails to non-responders etc. No non-invited responses allowed.

David Brake

Interviewee characteristics

• evenly split by gender

• 74% white

• education level from vocational to postgrad degree (73.9% were degree educated)

• 56% said they were 'middle class', 13% (3) self-identified as 'working class').

• Ages collected as ranges and participants were from all of the age brackets from 16-24 to 50-64 (though there were only two interviewees in the latter age bracket and no questionnaire respondents over the age of 65).

David Brake

Interviewee characteristics (cont.)

• Also included range of scores on:• “On a scale of one to five how well would a

regular reader of your weblog get to know you?”• “Roughly how many people do you think visit

your weblog in an average week? (Please guess even if you don't know)”

• “how important to you are the comments you get from readers?”

David Brake

Benefits

• Was able to get sample that varied around key variables of theoretical interest (eg number of perceived readers, degree of openness)

David Brake

Weaknesses I - social composition

• Difficult to recruit truly ‘disadvantaged’ bloggers (unclear how many existed then)

• Three “working class” people interviewed were a software engineer, an accountant and a “writer” (actually student)

• Of 6 without degrees, two were students, two were in mid-level IT, one accountant and one administrator (but self-identified middle class)

• Only two respondents combining working class, low educ and blue collar occupation - neither agreed to interview

David Brake

Weakness IIUse of search engine to construct

sampling frame

• Coverage and result presentation algorithm not well understood

• Possible bias towards more popular sites (though attempt made to compensate)

• Relatively clumsy (would be even harder if you wanted to draw sample across wide variety of weblogging platforms).

David Brake

Possible solutions

• Chyng Yang Jang ([email protected]) has a tool to automate extraction of Blogger “random blogs” (which still favours recently updated blogs)

• Could have “filled in” sample through snowball or cluster sampling (biased but better than nothing?)

David Brake

Further Questions? Comments?

Contact details:David BrakePhD researcher, Media and Comms department, London

School of Economicshttp://davidbrake.org/http://blog.org/

Thank you for your attention!