Does Class Exist?

In Dennis Condron’s new ASR article on academic achievement he writes:

“The second contribution [of this article] lies in conceptualizing and analyzing social class rather than socioeconomic status (SES).  This reflects the view that children growing up in different positions in the stratification hierarchy have categorically unequal and qualitatively different (rather than continuously graded) life and educational experiences.”

What evidence exists for this claim?  To me, this is the sort of claim that the paper should explore empirically, but let’s go ahead and see how he measures social class:

“I [Condron] code children as middle/upper class if either parent has a bachelor’s degree or higher or works in an executive, administrative, or managerial position and their household income is above the federal poverty line.  Children are coded working class if both parents have less than a bachelor’s degree and do not work in an executive, administrative, or managerial position and their household income is above the poverty line.  Finally, I code students as poor if their household income is below the poverty line, regardless of parents’ education levels and job positions.”

Then he examines the relationship between the class categories he created and academic achievement.

The standard objection to this approach is that it is throwing away data, the class categories hide important differences.  Whether you have one or two parents with a B.A. does not enter the analysis.  If your parents’ income is under the poverty line, their education is ignored.  Income differences above the poverty line are ignored, etc.

If I reanalyzed the data using parental income, education, and occupation, instead of class, I would be able to explain more variation in academic achievement.  There are times when I am more willing to “throw away” data, for example when I’m desperate for more degrees of freedom, but even then I feel bad about it.

This provocative blog post title: “Does class exist?” drew inspiration from Daniel Little.  I’m curious to see if it attracts more clicks.  It would probably be more to the point to ask whether the concept of discrete social classes is useful.  If it does not help one predict outcomes of interest, then I think not.


19 Responses to Does Class Exist?

  1. eli says:

    It strikes me as deeply politically naive to argue that the definition of social class is an apolitical question that can just be settled by purported disinterested social scientists as a question of analytical utility. The American fiction that social class doesn’t exist can be a powerful justification for maintaining social inequality, and the fact that there are a lot of people in the middle shouldn’t, in my view, in any way obscure the radical cultural, economic, and political disjunctures between different social classes in the US. In some analytic contexts, perhaps like the one you’re examining here, the margins of a social class can be blurry, yes. There isn’t necessarily one single measurable variable that indicates someone’s class standing; even something like wealth isn’t always a reliable indicator (upper-class people can wind up broke but don’t thus automatically become working-class). But the fact of blurry edges doesn’t obliterate the substantial commonalities of a social group. And sometimes class boundaries can be culturally maintained even regardless of the economic or educational commonalities that cross class lines (boundary maintenance is an important social process).

    I’m not going to venture a definition of class right here, and I think one would have to be very holistic about doing so, but if you take any uchicago professor and put them next to the guys who hang out next to the garfield green line stop, I think the (discontinuous) class difference between them would be very obvious to even the most naive observer and would be highly manifest in the social interaction (or rather avoidance) that would probably ensue. There has been some interesting writing about working-class people who go into academia and face immense class barriers and prejudice that might provide some useful illustration of these kinds of divides, for example:

    Anyway, the point is just that, yes of course nondiscontinuous class phenomena can be found, but that in no way prevents the continued maintenance of a class system.

    • quinn says:

      Hmmm I am not convinced by his definition of social class. Yes of course it’s a very measurable way of defining social categories through the means of education and poverty. However it does not take into consideration etiquette, social expectations, wealth, measurements of wealth, interactive styles, types of disclosure, disclosure rules etc. Class is defined by education and poverty and yet the social element of how it is defined is somehow discounted?
      BA degrees, managerial positions have a difference in credibility; in academic validity salary, opportunity,experience and peer community and therefore determine access and opportunity to self and environmental improvement.I am by definition a middle class child, growing up in a working class community. I talk, walk, dress, like a working class person. I have a loudness in my voice that has been attributed to a working class person by a middle class person and have thus been rejected by middle class individuals. I therefore consider myself to be working class and in degree education at the moment experiencing those class barriers and prejudices that you have acknowledged.

  2. dadakim says:

    does the ASR require authors to make data available for replication purposes? if so, seems like the perfect 2nd-year graduate student project.

  3. Michael Bishop says:

    Dadakim, The data in question is the ECLS-K, so one could apply for access. ASR does not require submission of code so one would have to ask Prof. Condron directly. See Jeremy Freese on this issue here: and

  4. jimi adams says:

    While it’s plausible that

    If I reanalyzed the data using parental income, education, and occupation, instead of class, I would be able to explain more variation in academic achievement.

    that’s actually an empirical question, no? It may not be the case. Let’s ask Dennis if he did something along those lines in alternate analyses (since i don’t see a note along those lines in the paper anywhere).

    dk – You and i have had this discussion offline (i think), but why not reiterate it publicly. While i am conceptually all for the “code/data publishing as appendices” move that you (and Jeremy) support, as Michael mentioned, it hasn’t made it to sociology, just yet. And, i have to admit that i am at least a little terrified of what this is going to mean/look like in practice. Can you political scientists be sure to work out the kinks quickly before we sociologists do adopt it? KThx!

  5. dadakim says:

    we political scientists haven’t worked it out, but like Freese, I think that’s the way to go. the journal in our discipline with the highest citation rate requires it — though this is not the flagship general interest journal of our discipline, rather a very methods-y journal, with a great deal of self-selection in the sample of authors there.

    in my 2nd year of quant methods at ucla, it was actually a required project to solicit an author of a published article for data and replication files and then replicate their findings, as well as incorporate any new analysis you thought they overlooked. i think it was a great way to train data users on how to be crystal clear about your methods (not just in the paper but in your analysis: name your variables, write extensive do-files with lots of comments; and in the post-mortem: post your data to an archive, i.e. dataverse) to avoid getting shot down by some snot-nosed grad student who just learned a likelihood function. 😉

  6. […] in publishing In the comments on a previous post, dadakim raises a pertinent question about publishing practices that hasn’t (yet?) been adopted in sociology (other than by SMR*, […]

  7. Michael Bishop says:

    Nothing I have said implies that inequality in income, education, friends, musical ability or physical attractiveness are unimportant. If Condron lacked data on income, education, and occupation but had data on social class then I wouldn’t object to his analysis. It is only because he chose not to use data which appears to be better according to my proposed objective standard.

    I am willing to grant that it is sometimes worth theorizing about something one cannot measure, but one must attempt to specify its relationship to things which one can measure, or will be able to measure. Otherwise the theorist is too little constrained by things a community of scientists can agree on. Politics may be, in some sense, inescapable, but if we abandon the attempt to relate our theories to what can be observed, we are giving the theorist the maximum amount of room to be ideological.

    The same criteria I use for deciding whether the concept of discrete social classes is useful can be applied to whether other concepts are useful: mathematics knowledge, athleticism, race, weight/bmi/obesity. I’m always open to someone showing me data in which more fine grained concepts/measurements make my original concepts/measurements unhelpful.

    Of course, this is all related to Popper… other things being equal, we should prefer theories which provide testable hypotheses.

  8. Michael Bishop says:

    Jimi, I agree that it is an empirical question, but it is one I am basically willing to bet my life on. I don’t know how much additional variation I could explain, but I could explain more. I would allow for non-linear effects in education and income, interactions, etc.

    Of course, if the additional variation I explain is truly tiny, one could argue that creating social class categories is justified because it makes the regression output easier to read. But this is not the grounds on which the decision is defended in the paper.

    I should add that despite the questions I raise, I think the paper makes some some important contributions to an incredibly important research program.

  9. Dennis Condron says:

    The class debate is a huge one, and settling that debate was not a goal of my study. With that in mind, I just want to offer two comments:

    First, the original post asks what evidence exists for the claim that “children growing up in different positions in the stratification hierarchy have categorically unequal and qualitatively different (rather than continuously graded) life and educational experiences.” The answer lies in the sentences immediately following that one in the article (p. 685), which include references to leading scholars in the area.

    Second, I was very thorough and careful to ensure that my findings did not depend on how I chose to operationalize social class. I report this in the article on p. 700 — I get the same pattern of results in supplemental analyses that use the continuous SES composite instead of my social class indicators.

  10. Michael Bishop says:

    Dennis, how generous of you to reply to my comments. A public dialogue between scholars is, in my mind, the very highest goal of a blog like this.

    You are certainly in the company of good scholars in arguing that social class is best conceptualized as a discrete variable, but the arguments I am familiar with, by authors you cite, do not convince me. Lareau & Cookson & Persell begin their empirical research by seeking out groups which are different from each other. Naturally this lends itself to seeing their differences as categorical, rather than continuous. I’m willing to accept that the idea of discrete social classes may be a useful simplification in certain analytic contexts, but I maintain that scholars often attribute class more reality than it has, the same way some economists might attribute the rational actor more reality than it has. Luckily, in your work and that of Lareau, and others who I feel overemphasize categories, there is still an immense amount for me to learn.

    I should have mentioned your note on p. 700 about alternate specifications, though I would have preferred to see specifications with parental education, income, and occupation each added separately.

  11. […] Frog Pond Effect in Schools Appearing in the same issue of the ASR as the Condron article I previously discussed, Robert Crosnoe publishes evidence that lower income students suffer some negative academic and […]

  12. Ned Naborsky says:

    Hi, I just happened across this blog while doing a search for ‘class’. I am very interested in the term ‘class’ and the concept behind it. When I try to define it in a way explainable in lay terms to other people I get bogged down. My own understanding of it comes mainly from Marx and Weber. Every attempt at thinking it through in order to explain it (even to mysef) results in a hopeless mess of open ends. It’s now at a point where I’m beginning to question its very existence. Hence I’m glad I stumbled onto this blog as I haven’t been able to find any coherent analysis of class on the internet so far.

    Could someone point me to any literature (as mentioned here) that discusses both why and why not ‘class’ is a useful/useless concept? Is class a concrete ontological reality or merely a social construction or an artifact of the analytical gaze? I want to understand both sides of the argument. Please fee free to refer to either sociological or political papers or books.

    Thanks to anyone who could help me with this.

  13. Hi Ned,
    Since Mike made this post I’ve stumbled on the work of Erik Olin Wright, who seems to be an especially prominent sociologist trying to retheorize a marxist analysis of class structure in more methodologically sophisticated ways. I strongly recommend you check out his website: it has tons of articles and you can even download the texts of his (several) books on class. If you want to get a coherent analysis of class, you could do worse than to start there.

    About your more concrete question, it seems to me that the adjective “useless” is itself useless without reference to some context of use. In other words, certain formulations of the idea of “class” may be more or less useful for certain research (or political) projects, but it makes no sense to me to ask in the abstract if “class” is a useful category. Useful for what?

    In a broader theoretical context, it seems to me that a theory of classes is ultimately among other things a theory of group formation and social differentiation. So insofar as one is interested in an analysis of the general social structure of contemporary society, a denial of the existence of social classes necessarily raises the further question: what alternative theory of group integration and differentiation is being proposed instead? (I suppose there may be methodological individualists so extreme that they don’t believe in any account of social groups at all, but that seems to me quite baffling.) I’m guessing that someone like Mike might argue that educational and racial groups are much more consequential than classes, but he’d have to comment on this himself.

    One last point. Your last question has to do with whether class is a “concrete ontological reality or merely a social construction or an artifact of the analytical gaze.” I rather like a 1987 paper by Pierre Bourdieu, “What makes a social class? On the theoretical and practical existence of groups,” that argues that it is all three of these at once and that there is no particular contradiction between them. It’s in the Berkeley Journal of Sociology vol 23. The argument, in essence, is that there are ‘objective’ groups by people who share similar characteristics in social space, that these ‘objective’ groups are only analytic constructs, but that people in a given ‘objective’ group can, by a process that Bourdieu calls ‘symbolic construction,’ form a concept of themselves as a group and hence become a class in practical and subjective terms. (In marxist terms, if this helps, Bourdieu is arguing that the passage between a class ‘in itself’ and a class ‘for itself’ is more contingent than marxists would normally admit. But you have to read the article yourself.)

    As for the argument against class, which is different in quantitative sociology than it would be in cultural anthropology (my own field), I predict that Mike could provide some references!

  14. Ned and Eli, Thank you both for your contributions. I strongly agree with Eli when he says:

    ‘ In other words, certain formulations of the idea of “class” may be more or less useful for certain research (or political) projects, but it makes no sense to me to ask in the abstract if “class” is a useful category. Useful for what?’

    The point of my original post was methodological and not specific to the concept of class (don’t discretize continuous data without good reason). But in certain circumstances I can imagine finding class a useful concept. I don’t particularly want to rank race, class, or education levels in order of importance in the abstract. Both the data, and the research question will bear on what concepts are most useful. Let’s take specific examples… what questions are we trying to answer?

  15. Ned Naborsky says:

    Thanks Eli and Michael. I will check out Erik Olin Wright’s work. It look’s like it’s along the lines of what I am thinking about.

    Obviously class does not exist in the abstract, as does little else, except maybe Euclidean geometry. By useful, I mean useful within its language game of conveying meaning about inequality in modern societies. When someone uses the term, what are they trying to convey? No matter the context, what do people want to say about the world? Is class a useful category? For example, an adolescent with a lemonade stand has the same ‘relation to production’ (Marxian use of ‘class’)as Rupert Murdoch. But so what? What is important about the grouping in this economic sense?

    I could be mistaken but class seems to run into some of the same problems as other ‘perilous ideas’ (as anthropologist Eric Wolf put it) as race, culture and ‘people’. I find that in studying political discourse class tends to get reified into a natural kind, putatively describing something real about the world. Obviously inequality and hierarchy exist. But is sorting people into ‘classes’ a useful way of understanding these things?

    Since I am not as familiar with in-depth sociological work on class, I don’t want to rule out the potential utility of class in this field. It was just intriguing to stumble across the provocatibve title which triggered so many things I personally have been working on and thinking about.

    Thanks for the references so far ( I almost forgot about Bourdieu’s work on class distinctions). I appreciate your feedback and welcome further suggestions for reading.

    (Sorry to confess my ignorance here but it’s difficult to tell: is this Michael Bishop’s blog?)

  16. Ned, this is a group blog which is sponsored by the Mathematical Sociology section of the American Sociological Association. Organized by Matthew Brashears, the other primary contributors have been Jimi Adams, Scott Golder, and myself.

    I don’t know whether I should be happy about having chosen a provocative title, which stimulates conversation, or ashamed, because it hints at grandeur, when the message that I think is more appropriate is humbler.

    Let me say this about the concept of class. The world is complicated, and there are advantages to working with concepts that one can measure because then you can look to see exactly what parts of empirical reality your theory explains and what parts it doesn’t. Now this is a simplification. Often the thing you can measure is only a proxy for what you’re truly interested in (what the theory is about). I might argue that a certain curriculum better teaches certain mathematical concepts that are necessary for solving certain real world problems, but its often the case that I only have one or a small number of tests that attempt to measure that mathematical understanding. I need to be worried that the curriculum improved the test performance without improving the conceptual understanding or the real world application. Concepts like measurement error, construct reliability and validity, and latent variables are relevant here. At the next level up (though they kind blur together), a particular dataset, you can think about which statistical models best illuminate which aspects of the data. The next level might be how that data and those models generalize to other contexts.

    Each time we jump a level of analysis we take a leap of faith that none of the new assumptions are problematic. All levels should be subject to scrutiny, but in practice, we specialize in analyzing certain levels and disciplinary standards and boundaries encourage this specialization.

    What is a useful concept? One thing that is useful is prediction. In what context does the concept improve prediction? I’m not saying that this should be the only standard, but it is an important standard.

  17. Mike Sobocinski says:

    I have found that in order to make references to “class” in a more rigorous and defensible way (when discussing anything beyond a simple correlation between some readily quantifiable proxy for class and another variable), that it is important to distinguish between individuals’ status sets and class as a collective category. (This jargon is straight from introductory sociology textbooks, by the way, which should be referred to if readers here are from other fields.)

    “Social class” is understandably confusing if/when it sets up an expectation that its aspects are all coherently expressed in neat categorical packages (e.g. lower class, working class, middle class, upper class). This is where SES *should* or could provide some helpful clarification – two particular attributes such as education and income, which are readily measurable, get offered as a simplified yet easy-to-understand concept. Even just with two variables, mismatches are notable and raise questions about the coherence of the categorical class concept. One solution might involve dealing with these types of continuous variables as explicitly separable factors within an analysis, even though they tend to correlate with each other. This seems more proper than to take people whose statuses on multiple variables are quite diverse and often divergent, and then lump into artificial class categories and *then* do an analysis. Class is often just a shorthand concept, yet one of the popular textbooks on social classes (was it Kerbo, Gilbert and Kahl, or both?) points out that status inconsistency is one of the defining characteristics of the middle class – a mismatch between education and wealth, for example, in which college professors are the most highly educated in society but are not the wealthiest members, or between prestige and power, occupation and income, and so on.

    For many purposes, it is neither necessary nor helpful to invoke a broad, categorical concept of class, when relationships involving specific measurable variables are much clearer and the data more readily available.

  18. This is a message to the admin. I came to your Does Class Exist? | Permutations page by searching on Google but it was difficult to find as you were not on the front page of search results. I know you could have more visitors to your website. I have found a website which offers to dramatically increase your rankings and traffic to your website: I managed to get close to 500 visitors/day using their services, you could also get a lot more targeted visitors from Google than you have now. Their services brought significantly more traffic to my site. I hope this helps!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: