Classes and classification: comment on Chan and Flemmen, Jarness and Roselund

Key takeaways

content: "@lizardoClassesClassificationComment2019" -file:@lizardoClassesClassificationComment2019

Reading notes

Imported on 2024-06-10 13:11

⭐ Important

• & However, in this case, we can take a lesson from Durkheim. There’s nothing sociologically special about using ‘social class’ defined, as Chan does, as a ‘structure of inequality that is rooted in the social relations of economic life, that is, [RE]lations in [LA]bour [MA]rkets and [PRO]duction units’ (Chan 2018: 23) to split people into groups-on-paper. Chan, Goldthorpe (and others) call this ‘social class’, but we might as well call it RELAMAPRO. So we have people who fall into different RELAMAPRO classes, and we have a bunch of research (nicely reviewed by Chan in his response to FJR) showing that assignment to RELAMAPRO classes predicts all kinds of outcomes. (p. 907)
• & In this respect, as Weeden and Grusky have convincingly shown in a series of papers (e.g., Grusky and Weeden 2001; Weeden and Grusky 2005; Weeden and Grusky 2012), the worst thing American stratification researchers ever did was to lump together these really nice, socially and phenomenologically valid classes, the closest to Durkheimian totemic groups we have in the contemporary world (Martin 2000), to create little 5 by 5 tables featuring such meaningless lumps as ‘Managerial/Professional’ all so they could fit a log-linear model and get an L-squared statistic. (p. 908)
• & Multidimensional scaling (MDS) of the distance matrix implied by such a cross-classification will yield a sizeable first eigenvalue,2 and the eigenvector corresponding to this eigenvalue can be used to ‘rank’ the occupational groups. Chan and Goldthorpe like to call this class assignment ‘status’ (because fancy occupations end up on one side and less fancy ones on the other) and treat it as a continuous variable.3 Chan also tells us that Ed Laumann says it is OK to call this ‘status’. (p. 908)
• & However, we might as well call it schmatus, since ignoring higher-order dimensions was an arbitrary choice and the theoretical interpretation of the first set of eigenvalues as pointing to relations of ‘hierarchy’, ‘equality’, or what have you is just that; an interpretation as is noted by FJR.4 (p. 908)
• & The point is that the schmatus way is just a different way of sorting people into classes and is no more or less appropriate than the RELAMAPRO way. (p. 909)
• & What FJR should have done is that rather than fight Chan/Goldthorpe in the realm of ‘conceptual distinctions’ between two (why only two?) ways of lumping people into classes, they should have brought the fight to the realm of strategies of inquiry, in particular methods designed for what computer scientists (without having read Durkheim or Bourdieu) call classification, (of which MCA is one of the many options but the SVD is always your friend (Martin and Porter 2012)) because ultimately we are talking about whether net effects versus some other approach actually at tuned to the classificatory job is the way to go (p. 911)
• & they cannot meet Chan’s (regression-based) pleads for ‘falsifiability’, ‘testing of competing arguments’, and ‘evaluation’ (a game people in the US played in the 1980s and 1990s visà-vis Bourdieu with relatively little to show for it). (p. 912)