Class, status and lifestyle: on omnivores, distinction, and the measurement of social position

Key takeaways

content: "@lambertClassStatusLifestyle2019" -file:@lambertClassStatusLifestyle2019

Reading notes

Imported on 2024-06-10 13:10

⭐ Important

• & Yet while Chan’s analysis is clear, logical and largely persuasive, it includes several empirical strategies that I suspect may be sub-optimal. As Flemmen et al. highlight, ‘status’ might be related to many more things than education and occupation, so Chan’s use of these items to assess status consciousness might underestimate status consciousness. (p. 887)
• & More generally I distrust the attempt (p. 887)
• & to meaningfully disentangle the influence of several linked explanatory variables (income, education, ‘class’ and ‘status’) and I distrust the optimality of the tripartite categorization of omnivorousness (which seems more naturally a continuous dimension – if so, measuring its features through a categorization might diminish the accuracy of the relevant estimated associations). (p. 888)
• & Chan’s analysis does not include controls for two other things which might well account for another small part of the patterns observed – namely social origins, and locality or micro-geography. (p. 888)
• & As Chan indeed acknowledges, multivariate modelling methods will in many circumstances provide us with lower-bounds to an estimate of a relevant correlation/association, ultimately due to measurement error and omitted variable bias. I suspect as above that the patterns that are reported are attenuated a little bit more in this direction than is recognized. (p. 888)
• & Flemmen et al.’s fascinating analysis is not as parsimonious as Chan’s! (p. 888)
• & Like Chan, Flemmen et al. also use data reduction techniques, but in a manner that is much more central to the conclusions of the paper, and they supplement their interpretations with descriptive and interpretive accounts (rather than by contextualizing patterns through multivariate models as does Chan). (p. 889)
• & I’m convinced that Flemmen et al. are correct in their argument that social interaction distance dimensions don’t translate to ‘status’, and that ‘class’ and ‘status’ are not so easily disentangled (in other publications with colleagues I have made very similar arguments). Flemmen et al. reason that ‘class’ and ‘status’ could overlap and their roles should be understood in a more integrated analysis of the entwinement of the social and lifestyle space (which they proceed to conduct). I’m persuaded by Flemmen et al.’s thoughtful argument, which offers a strategy for studying social position effectively without risking unrealistic claims about our ability to disentangle it into different dimensions. (p. 889)
• & My main quibble with Flemmen et al.’s analysis is a concern that their interpretation of data reduction patterns may be unduly deterministic and potentially biased. For example, in Flemmen et al.’s empirical lifestyle dimensions I simply saw ‘age’ and ‘gender’ in dimensions 2 and 3, but the authors’ subjective interpretation of these was more nuanced, namely as dimensions discriminating by playfulness and by intellectualism. (p. 889)
• & Analytically I increasingly think that the main weakness of techniques like correspondence analysis and clustering algorithms are that they make it hard to effectively express statistical uncertainty around its conclusions (compare the fluency with which we can interpret the confidence intervals portrayed by Chan). (p. 889)
• & In any case, the methodological risk of accepting categorical divisions without ambiguity is of translating patterns from weak to strong (to focus on categories rather than individuals might invite an ecological fallacy: correlations between aggregate units will often be different – typically stronger – than those at the microlevel – so a narrative about aggregate units might imply tighter patterns of (p. 889)
• & association than is appropriate). (p. 890)