Dear Irv,

I'd like to think through the gap issue a bit, as I'm not sure that just eliminating data is the right way to go. Gaps in data only work if they correspond to gaps in real life. The line between classes is not as clear as, say, the lines between "Catholics" and "Protestants" or "Female" and "Male", because pretty much everyone uses these labels in the same way. (My current students, though, don't know what a "Protestant" is, because they were raised with the term "Christian".)

However, it's not clear to me that Upper-Middle / Middle-Middle / Lower-Middle / Upper-Working / etc. form an unproblematic continuum. That's in part because there's no universal definition of what "middle" and "working" mean anymore. That system worked reasonably well in the 1930s, but it doesn't clearly describe the present economy. So you may have asked folks to create more distinctions than they make in ordinary life.

I suggest that you run your data in a couple of ways, one of which you've done by creating a gap, and the other of which involves collapsing your groups into "Middle" and "Working". This may tap ideology as much as it taps actual class position; "working" implies trade-unions, resistance to exploitation, sweat-of-the-brow stuff, and "middle" implies more security, professionalism, family-centered normality, etc. But tapping into this ideology might be useful for two reasons. First, you're looking for ideological and behavioral correlates to class position, so dividing the classes by such root identifications is likely a good idea. Second, you'll be using the same categories that Basil Bernstein's used in his research on British language use. That research would suggest that political discussion including kids fits a middle-class elaborated speech model, while lack of discussion or excluding kids fits a positional-speech model.

Best,
Jim

ADDED PARAGRAPH (I ran over the 2000-character limit):

For what it's worth, the 2006 General Social Survey asked people to assign themselves to four classes: "Lower" (6.4%), "Working" (45.3%), "Middle" (44.5%), and "Upper" (3.2%). (The rest were "Don't Know" or "Refused to Answer"). Did they assign themselves right? Not likely. But do the differences say anything about them that's useful? Probably. (Significantly more "Working" than "Middle" said spanking kids is okay; no difference on attitudes toward gay marriage, on whether men are better suited than women for politics, etc.)

You can find the GSS online at http://sda.berkeley.edu/archive.htm The site includes software for running on-line cross-tabulations.

thanks, jim. always good to hear from you. I'm not pretending this is a well-controlled study. I was just curious about the question as a consequence of something I was writing about social class and argument. I really just appropriated labels that other stratification theorists have used (howard kerbo, one of my key sources). any of us who write about class no that the labels are freighted and that self-assignment is far from innocent. nevertheless, most of the people i have sent this out to will probably have thought about this issue a bit (which is why they may have decided to help me out with the question). Because of the play in categories, i wouldn't want to assign too much weight to differences between adjacent categories--but at the upper an lower reaches, the information reaches significance--or at least implications are there. In fact, most of the literature on social class differences implies differences in the kind & degree of political discussions, but I couldn't find anyone who had actually done any research on the specific question (bourdiieu has a few paragraphs that address the differences in social classes and whether people have political opinions on foreign policy); hence the survey. The information does substantiate (and even gives me percentages) of what other conversations on social class and discourse habits imply. thanks for that reference. i'll be able to use it. i will run the data by collapsing. definitely.
irv

I think the only way to go over the 2000 character limit is for the writer to create a separate workspace page. I'm not sure how one does that.