Do Big Classes Really Matter ?

One of the more fashionable claims in contemporary sociology is that individuals’ life chances, attitudes, and behaviors have become decoupled from their class locations. The few empirical analyses relevant to this claim inevitably begin with nominal conceptions of class that aggregate occupations into “big classes” or reduce them to scores on vertical scales of prestige, socioeconomic standing, or cultural or economic capital. These conventional practices strengthen the postmodernist critique of class models by obscuring the local subcultures that emerge at the detailed occupation level and by otherwise concealing the proximate mechanisms that link structural locations with lifestyles, life chances, and other individual-level correlates of class. This article examines the explanatory losses incurred by aggregating or scaling occupations for a diverse set of 50 outcomes culled from the Current Population Survey (CPS) and the General Social Survey (GSS). The results show that conventional class analytic approaches account for only a modest proportion of the total bivariate association between the site of production and life chances, attitudes, and behaviors. Moreover, there is nearly as much heterogeneity within big classes as between them, implying that big class models fail to capture much of the structure at the site of production. Conventional class models also fare poorly in illustrative multivariate analyses, thus undermining the fallback position that big-class heterogeneity disappears when standard demographic variables are controlled. It follows that sociologists who wish to understand the structural sources and correlates of micro-level behavior and attitudes would be well served to take the division of labor seriously. The class-analytic tradition has a long and volatile history in which waves of creative scholarship have been interspersed with periods of considerable cynicism. The current cynical phase has been ascendant for nearly two decades, thereby suggesting that something more than academic fad may be at work. This ongoing dissatisfaction with class models is most prominently articulated in the claim that lifestyles, political behavior, and consumption practices are a function of individual tastes and choices that are themselves largely unrelated to class membership (e.g., Clark and Lipset 2001; Beck 2000; Kingston 2000; Inglehart 1997; Pakulski and Waters 1996). The resulting individualization of inequality implies that the site of production has lost much of its hold on individual-level outcomes. The net cast by this “postmodernist” critique of class analysis is wide, but its primary targets are categorical models of class that represent the stratification system in terms of highly aggregate groupings (“big classes”) and gradational models that map disaggregated occupations onto vertical scales of socioeconomic status (Hauser and Warren 1997), prestige (Nakao and Treas 1994), or cultural and economic capital (e.g., Bourdieu 1984). In responding to this critique, some proponents of class analysis have simply defended their favored specification, whereas others have sought modest and incremental revisions to them. As a result, although there is considerable debate over how big classes should be defined (e.g., Evans and Mills 1998), how occupations should be scaled (e.g., Hauser and Warren 1997), or whether an aggregate or gradational approach is preferable (e.g., Rytina 2000; Hout and Hauser 1992), proponents on all sides appear content to assume that either the aggregate or gradational approach should be adopted in some form. We challenge this assumption by offering a theoretically and empirically viable alternative to aggregate and gradational class approaches. This approach rests on a new map of the site of production that reflects the social groupings that emerge around functional niches in the division of labor. Rather than reducing occupations to gradational scales or aggregating them into big classes, our approach relies on realist categories that, unlike the nominal units of conventional class analysis, are embedded in the fabric of society. As we have argued elsewhere, occupations are deeply institutionalized social groupings 1 that can act collectively to defend jurisdictional claims over tasks and alter the conditions of work for their members, define lifestyles and consumption practices that are binding on members, encourage social interaction with like-minded individuals, and provide subjectively meaningful categories with which employers, workers, and the lay public perceive themselves and others (Grusky and Weeden 2001, 2002). We argue that this disaggregate approach is better suited than conventional models for assessing and understanding the correspondence between objective conditions at the site of production and individual-level “outcomes” (e.g., attitudes, life chances, political behaviors). The reasoning here is simple: Namely, big-class models falter because the mechanisms that generate homogeneous outcomes require institutionalized categories that are typically found only at the level of detailed occupations, while gradational representations of class falter because these mechanisms play out differentially across detailed occupations with similar vertical standing. By treating occupations as purely nominal categories, big class and gradational approaches fail to capture the structure generated at the detailed occupation level, thereby leaving the class-analytic enterprise open to postmodernist critique. If class models are instead based on institutionalized boundaries at the site of production, we suspect that the correspondence between productive locations and various outcomes will become strong enough to cast doubt on the scholarly retreat from class analysis. At the very least, it is worth examining this possibility rather than simply making do with conventional approaches or, worse yet, abandoning all class models in favor of some post-structuralist alternative. We proceed by quantifying the explanatory losses that are incurred in conventional sociological analyses that aggregate or dimensionalize detailed occupations. We assess these losses by examining the association between competing maps of the site of production (i.e., big classes, scales, and occupations) and a range of possible outcomes. For purposes of presentation, we group these outcomes into six topical domains: life chances (e.g., income, education), demographic structuration (e.g., segregation by race, ethnicity), political attitudes and behaviors (e.g., party identification, political ideology), social 2 attitudes and dispositions (e.g., tolerance, racial prejudice, sexism, anomie), institutional participation (e.g., marital behavior, religious attendance), and lifestyles and consumption practices (e.g., television use, association memberships). These domains cover the wide diversity of research interests in the discipline, not just those outcomes over which stratification scholars typically obsess. As we will show, the costs of aggregation are greater in some domains than in others, but in all domains they are considerable. Does it Matter Whether Class Matters? Anti-class rhetoric has so permeated the discipline that it is now necessary to argue what once was taken for granted: that sociologists of all stripes, not just self-identified class analysts, should care whether conventional class models adequately capture the association between productive locations and individual-level outcomes. Therefore, we begin by briefly reviewing the main sources of sociological interest in the class-outcomes relationship. As laid out below, this relationship is directly relevant to at least four research traditions, all of which have historically conditioned on aggregate class models or vertical scales. These traditions involve (a) constructing new class maps in which the dividing lines are explicitly drawn to best capture individual-level heterogeneity in life chances, lifestyles, or other attitudes or behaviors, (b) determining the extent to which class maps that are predefined in terms of exploitation, working conditions, or other conventional criteria account for individual-level outcomes and are accordingly high in “proximate structuration” (Giddens 1973), (c) examining the net effects of social class in the context of full multivariate causal models that purge out all effects (e.g., race, religion) that are correlated with class, and (d) constructing multivariate models of various outcomes (e.g., voting behavior, social attitudes) in which a class measure is regarded as a statistical “control” rather than a focus of inquiry in its own right. In all of these traditions, aggregate class models or scales have invariably been applied; as a result, the extent to which these traditions have, by virtue of such aggregation or scaling, misrepresented the associations and effects of interest is simply not known.