Abstract
When a group shares a viewpoint on a status order, their consensus imparts legitimacy to their shared understanding of that order. Conversely, a group espousing multiple viewpoints undermines the notion that one “true” hierarchy exists. To build empirical knowledge about how social groups contribute to the construction of status orders, we take the occupational hierarchy as a case study and map the structure of agreement across intersectional groups. First, we quantify the extent to which groups (1) agree internally on their occupational rankings (within-group consensus) and (2) agree with other groups (intergroup consensus). Using General Social Survey data on occupational perceptions, we find a cluster of privileged groups—namely, highly educated White men and women—who agree internally and with each other on the occupational status order. Lesser advantaged groups exhibit less internal agreement and do not cohere around an alternative conceptualization of value, leaving unchallenged the consensus of privileged groups.
Keywords
Consensus regarding a set of beliefs imparts legitimacy and authority to those beliefs (Johnson, Dowd, and Ridgeway 2006; Ridgeway and Correll 2006). For example, a group that perfectly agrees on how to rank a set of “social categories or ‘types’ of people” (Ridgeway 2014) by their relative standing is a group backing one ranking system. Without contestation or diversity in opinions, the agreed-on ranking becomes legitimated as social fact: “what is becomes what is right” (Johnson et al. 2006:57). In contrast, members of low-consensus groups may personally believe that “type” matters for evaluating a person’s worth, but collectively, a low consensus group does not build legitimacy around any one ranking system, thus undermining the conclusion that types carry inherent status meaning.
Despite theoretical clarity around the importance of consensus to legitimacy, few empirical studies seek to delineate the extent to which groups generate consensus around status orders. There is a rigorous line of research from status construction theory about how and under what conditions beliefs about worth can spread through interactional encounters to create consensus (see Ridgway and Correll 2006). This experimental tradition, however, is not designed to address the extent to which beliefs about worth actually become consensual in a population and if consensus is driven by particular groups.
Recent research on perceptions of the occupational hierarchy gives us strong reason to believe that groups “contribute” unequally to building legitimacy around status orders. Lynn and Ellerbach (2017), using the 1989 General Social Survey (GSS), discovered that respondents with high levels of education compared with lesser educated respondents showed significantly greater agreement in their ordering of a large set of occupational titles on a nine-rung “social standing” ladder. Valentino (2022) replicated this education finding with the 2012 GSS and reported a similar pattern for race (White respondents exhibited more consensus than non-White respondents), gender (men more than women), and earnings (high earners more than low earners). In sum, on each of these isolated dimensions, the more privileged category of respondents showed greater internal agreement, thus creating legitimacy for their shared view, while the less privileged groups held more diverse opinions about occupations.
These findings raise two questions. First, how does this group of background characteristics—education, race, and gender—intersect to shape how people evaluate occupational status? We know from standpoint theory (Harding 1991; Sweet 2020) and intersectional approaches more generally (Collins and Bilge 2016) that viewpoints are derived from living life at the intersection of interdependent systems of inequality. For example, the experience of someone who identifies as Black, female, and highly educated is not the mechanical average of those three identities but rather the unique product of that combined identity. Second, building from this line of inquiry on within-group agreement, what insights are revealed when we examine the extent to which groups agree internally alongside the extent to which groups agree with other groups? Are there clusters of groups who perceive the occupational order similarly, creating one or more “voting blocs” competing to legitimize their definition of the occupational hierarchy?
To address these questions, we conduct a new analysis in which we define groups by the intersection of multiple demographic characteristics and employ network methods to describe the formal structure of agreement around a perceived status order. We begin by quantifying within-group consensus, or the extent to which groups internally agree on the occupational order. Unlike Lynn and Ellerbach (2017) and Valentino (2022), however, we examine groups defined by multiple characteristics rather than single characteristics in isolation. Following Cech (2022), we divide our sample into mutually exclusive intersectional groups, some who enjoy a high degree of structural privilege (e.g., White men with graduate degrees), some who do not (e.g., non-White women without a high school diploma), and some who enjoy privilege on some dimensions but not others (e.g., Black women with graduate degrees; White men with a high school diploma). This allows us to compare how these dimensions combine to shape viewpoints and group consensus in nonadditive ways.
Second, we quantify between-group consensus, or the extent to which members of any two intersectional groups similarly rank occupations, which neither Lynn and Ellerbach (2017) nor Valentino (2022) explored. We then graph the structure of intergroup consensus layered with information about internal consensus. Our reanalysis of the 1989 and 2012 GSS occupational perceptions data shows a cluster of privileged intersectional groups displaying high internal and intergroup agreement; the remaining, lesser-privileged intersectional groups exhibit significantly less of both. 1 Overall, this pattern of consensus suggests that a privileged minority builds legitimacy for their version of the occupational hierarchy through (1) their own achievement of consensus and (2) the absence of organized contestation from the lesser privileged majority.
Within- and Between-Group Consensus on the Occupational Hierarchy
The social valuation of occupational roles is a long-standing area of sociological inquiry (e.g., Boltanski and Thévenot 1983; MacKinnon and Langford 1994; Martin 2000; Treiman 1977; Zhou 2005) that continues to capture scholarly attention (e.g., Freeland and Hoey 2018; Jiang 2023; Kmetty, Koltai, and Rudas 2021; Maloney 2020; Noonan, Lynn, and Walker 2020; Valentino 2021). A standard measurement strategy used to study this topic—employed in the 1989 and 2012 GSS—asks respondents to pile sort a set of occupations into a nine-rung ladder based on their perceived “social standing” with nine indicating the highest standing. These pile-sort results measure how individuals in different social locations articulate the “orders of worth” in which they are embedded (Boltanski and Thévenot 2006; Lamont 2000). We refer to an individual’s entire pile-sort outcome as their “viewpoint” or perception of the occupational status order.
Within-Group Consensus
Quantifying consensus around a set of beliefs is a cornerstone of measuring culture (Martin 2002; Wood 2023). With respect to perceived status orders, a group in which all members sort objects into precisely the same tiers exhibits classification consensus or perfect within-group agreement, an outcome implying that members espouse the same underlying valuation logic. 2 In Accominotti, Lynn, and Sauder’s (2022) terms, a group with perfect consensus constructs a “bright” as opposed to “blurry” hierarchy. Brightness is clarity regarding the “natural” relational structure of objects, that is, when “every unit in the hierarchy can be positioned unambiguously with respect to every other . . . [and units are] clearly of higher, lower, or equal status” (Accominotti et al. 2022:93). 3
In the present study, we examine within-group consensus for intersectional groups based on gender, race, and educational degree rather than single-characteristic groups. 4 These three variables are included near universally in analyses of social phenomenon because each dimension is theorized to impact how people think, how they are judged, and more generally, their life experiences (Collins and Bilge 2016; see also Monk 2022). Not surprisingly, these dimensions also appear to shape consensus around occupational status (Lynn and Ellerbach 2017; Valentino 2022). It is unclear, however, how these dimensions combine to produce consensus. For example, Black Americans might, on average, exhibit more heterogeneous pile sorts than White Americans (Valentino 2022) because there is greater heterogeneity among Black Americans of all educational levels. Conversely, educational attainment might shape how Black Americans perceive the status of occupations more so than White Americans, creating more subgroup homogeneity among Black Americans by education level but greater aggregate heterogeneity among Black Americans as a whole.
To our knowledge, there is not a line of research on why some intersectional groups are more likely to share one valuation viewpoint and other intersectional positions produce greater viewpoint diversity. Previous research, however, identifies three types of forces that could increase valuation homogeneity within groups. These forces are analytically distinct but likely intertwined in practice. First, beliefs about worth are more likely to be homogeneous within a group if members are exposed to a limited pool of ideas in general. For example, in the context of educational goals, Harding (2011:327) hypothesizes that disadvantaged neighborhoods are more likely to expose youth to a “heterogeneous array of cultural ideas” about schooling because such neighborhoods are highly diverse with respect to “occupational statuses, incomes, education levels, reliance on public assistance, [and] involvement in crime.” Second, valuation homogeneity is likely higher when groups are associated with institutions, organizations, or tightly connected networks that can socialize or discipline members into a certain way of thinking and provide them similar information that can be used in forming viewpoints (Zhou 2005). Wood (2023), for example, discusses the homogenizing effect of certain religions on parenting beliefs, and Harding (2011: 327) theorizes that wealthier neighborhoods are likely to use “social networks for common goals such as maintaining order” such that “alternative cultural orientations will have fewer opportunities for public expression.”
Third, valuation homogeneity likely increases when groups adopt black-white logics that reduce informational complexity. Lynn and Ellerbach (2017), for example, showed that those with more education exhibited greater consensus in their perceptions of the occupational hierarchy because they were more likely to adopt a simple, self-justifying “rule” when sorting occupations: “good” jobs require higher education, and “bad” jobs do not. The authors also speculated that the lack of consensus among the less educated could be due to exposure to multiple or competing institutional logics (e.g., religion, military), which could undermine the adoption of black-white ways of judging.
Generalizing from these findings, we suspect that those sitting in structurally and materially advantaged intersectional positions will exhibit greater consensus on occupational ratings because they are less likely to come from environments in which they are exposed to a large variety of “cultural orientations” and more likely motivated to organize around a viewpoint that preserves their advantage (Valentino 2022).
Between-Group Consensus
Whereas exploring consensus within groups can elucidate forces that homogenize viewpoints, exploring similarity between groups can clarify the structure of agreement for an entire collective. To illustrate, Figures 1a to 1d are network maps of four possible structures, each consisting of a collective with six hypothetical groups that we assume are equally sized and resourced for analytical simplicity. In each graph, a node represents an intersectional group, and its shade corresponds to the amount of within-group consensus on a status ordering task. Dark nodes are groups with a high degree of consensus, and light nodes are groups with a low degree of consensus. Two nodes are closely connected by a dark tie if intergroup agreement is high; two nodes are distantly connected by a light tie if intergroup agreement is low.
Four Stylized Agreement Structures (a) Global Agreement (b) Extreme Diversity (c) High-Consensus Core, Low-Consensus Periphery (d) Competing Viewpoints
Figure 1a thus represents global consensus: all six groups exhibit high within- and between-group consensus. Conversely, Figure 1b (all light nodes and ties) represents extreme viewpoint diversity—the kind of diversity expected if all gave random orderings. In both Figures 1a and 1b, groups are irrelevant to the construction of the status order. In Figure 1a, the average view is evenly backed by all the collective, whereas in Figure 1b, the average viewpoint is the mathematical aggregation of many different viewpoints and thus may reflect nobody’s actual viewpoint.
In Figure 1c, a strong consensus emerges among only half the population. Three groups converge on one understanding of hierarchy (i.e., the central cluster of dark nodes connected by dark ties), but the other three groups (light nodes and ties in the periphery) do not. In Figure 1c, consensus within the core and the lack thereof in the periphery leaves the core’s version of the hierarchy uncontested. Figure 1d illustrates a collective organized around two competing viewpoints. Each cluster in Figure 1d converges on an understanding of hierarchy, but their respective viewpoints share little in common.
In sum, Figure 1 illustrates distinct structural patterns of agreement, each of which has implications for how social inequality could be maintained and reproduced when considering that some groups enjoy more structural privilege than others. For example, in Figure 1c’s core-periphery pattern, the ideologically cohesive core is akin to a voting bloc that, by way of consensus, could drown out the voice of the periphery even though they each contain half the population. Resource differences between these groups—if the core comprised more privileged groups and the periphery comprised lesser privileged groups—could exacerbate this effect.
Data and Measures
We use the 1989 and 2012 occupational ratings data from the General Social Survey (GSS), which covered 740 and 860 occupational titles, respectively (for details, see Nakao and Treas 1994; Smith and Son 2014). Our analysis focuses on the set of 40 and set of 20 occupational titles that all respondents were asked to rate in 1989 and 2012, respectively.
Intersectional Groups
We divide respondents into intersectional groups (hereafter, “groups”) based on race (White, Black, other), sex (male, female), and highest degree attained (graduate, bachelor’s, junior college, high school, no diploma). This formally results in 30 groups, but not all have respondents. Also, we exclude intersectional groups with fewer than five respondents and those based on the “other” race category because of analytical ambiguity. 5
Within-Group Consensus
To calculate within-group consensus, 6 we begin by measuring pairwise distance, or the extent to which two respondents differ in their ratings of a set of 40 (1989 GSS) or 20 (2012 GSS) occupational titles on a scale of 1 through 9. We calculate pairwise distance using Manhattan distance, which is the sum of the absolute or raw difference between two vectors. This measure captures any type of deviation between two vectors. Two raters, for example, could order occupations in the same way but use different parts of the scale (e.g., 9, 8, 7, 6 vs. 5, 4, 3, 2). These two raters would be perfectly correlated using rank or Pearson correlation measures, but their Manhattan distance would be positive. When Manhattan distance is zero, two individuals sort occupations in exactly the same way.
Because a group is a collection of pairs, we construct a within-group consensus index capturing the extent to which all pairs in a group agree on average. A score of 1 corresponds to a group with perfect consensus, and a score of 0 is the amount of consensus expected if all members had rated occupations randomly. Put differently, a score of 0 is akin to maximum pile-sort randomness, and 1 is maximum pile-sort homogeneity. We find that observed within-group consensus levels range from about .0 to .5 in both 1989 and 2012. An index score of .50 corresponds to an average difference of about 1.5 ranks on each occupation. A high negative index score would indicate systematic disagreement within or between groups (e.g., one group assigned all 1s to every occupation, and another group assigned all 9s), but we do not observe such scores in either 1989 or 2012.
Between-Group Consensus
Agreement between two groups is calculated as the average pairwise distance of every unique, cross-group pair of respondents. Similar to within-group consensus, a score of 1 indicates that two groups are in perfect agreement with each other, and a score of 0 indicates that two groups share no more in common than what would be expected from random orderings. Intergroup consensus ranges from roughly 0 to .5 in both 1989 and 2012. 7
Results
Within-Group Consensus
Figure 2 summarizes the extent to which members of a group rated a set of occupations in the same way. To facilitate comparison of consensus levels across groups, we build 95% confidence intervals (bars) for each point estimate (triangles); group-specific standard errors are bootstrapped based on with-replacement resampling over 1,000 iterations. Similar to the color scheme in Figure 1, darker triangles correspond to higher consensus levels. Results are organized by education because degree induces far more spread in consensus relative to race and gender. 8
Within-Group Consensus (a) 1989 General Social Survey (b) 2012 General Social Survey
Figure 2 shows that most groups have consensus levels greater than zero, but groups vary significantly in their level of internal agreement. This pattern is clear even though some confidence intervals are wide due to small sample sizes. Overall, groups with more education tend to exhibit more internal agreement, but the homogenizing effect of education is inconsistent by sex and race within a given year and across 1989 and 2012, a result we return to in the discussion. 9
Between-Group Consensus
Intergroup consensus is also highly variable. Figure 3a plots the results of both internal and intergroup consensus for the 2012 GSS. Nodes in Figure 3a are intersectional groups, and both nodes and ties are shaded gradationally according to consensus level; darker nodes and ties indicate higher internal and intergroup agreement, respectively. 10 The pattern in Figure 3a resembles Figure 1c—there are darker nodes and ties in the core and lighter objects in the periphery. In Figure 3b, we highlight this core-periphery pattern by reducing Figure 3a to just two shades: dark nodes and ties are at or above the 80th percentile of their respective distributions, and light objects are below this threshold. Figure 3b reveals a subset of groups who share one understanding of the occupational hierarchy, whereas all remaining groups project a “blurrier” occupational order.
Within- and Between-Group Consensus, 2012 General Social Survey (a) Gradational Shading of Consensus Levels (b) Binary Shading of Consensus Levels
Unlike the toy example in Figure 1c, however, real-world groups differ in both size and structural advantage. To capture these group differences, Figure 4a replicates Figure 3b, but we now label nodes by their intersectional characteristics and size them by the percentage in the group with self-reported income over $90,000 (i.e., larger nodes are groups with higher incomes on average). Figure 4a clearly shows that groups in the core have characteristics associated with more privilege (e.g., White, highly educated, and higher income) relative to the periphery. 11 In contrast, Figure 4b replicates Figure 3b, but now nodal size is proportional to the numerical size of the group (i.e., larger nodes are groups with more respondents). 12 Figure 4b shows that the privileged core is a numerical minority. Together, Figures 4a and 4b show that consensus around the occupational hierarchy in 2012 is driven by a privileged minority who are “on the same page.” 13
Figure 3b with Nodal Characteristics, 2012 General Social Survey (a) Structural Advantage (b) Numerical Size
Discussion
Examining the structure of consensus in demographic and relational space reveals that groups contribute unequally to legitimizing the belief that occupational titles carry status meanings. A cluster of privileged intersectional groups—namely, White men and women with college and graduate degrees—agree on what constitutes high and low standing, but for lesser advantaged groups, beliefs about occupations are highly diverse. Lesser privileged groups do not cohere around an alternative conceptualization of occupational value (either internally or collectively) and thus do not challenge the consensus upheld by the privileged core. As predicted by legitimacy theory (Johnson et al. 2006), it appears that both the presence of consensus among higher status groups and the relative absence of coordinated resistance among lower status groups contribute to the validation of the beliefs held by higher status groups.
A comprehensive analysis of the content of consensus is beyond the scope of this study, but the current evidence strongly suggests that privileged groups are in agreement on protecting the status of occupations that require more education (see Lynn and Ellerbach 2017). For example, using the same measure of training time as Lynn and Ellerbach (2017), we find that the correlation between occupational ratings and educational requirements is .835 for White men with graduate degrees in 1989 but only .677 for White men without a high school diploma in 1989. In other words, the version of the occupational hierarchy that privileged groups back is one that favors themselves (see Valentino 2022).
Another important finding is that although consensus generally declines from most to least educated in both years, sex and race do not consistently generate more or less consensus within education levels or across years. This result underscores the importance of not treating intersectional groups, or social groups in general, as monolithic entities (see Monk 2022) and reiterates the importance of thinking more deeply about why social location may or may not lead to homogeneous views. That is, separate from asking why people in different intersectional positions might have different viewpoints on average, why do some intersectional positions generate more viewpoint heterogeneity than others? For example, why in 2012 are White women without a high school diploma more likely to converge on the same viewpoint compared with Black women without a diploma? Why do White women with graduate degrees replace White men with graduate degrees as the most homogeneous group from 1989 to 2012? To be clear, high versus low internal consensus does not imply that a group is skilled or unskilled at observing some underlying truth. Rather, consensus is high when the forces that homogenize are stronger than the forces that diversify group members’ opinions. Education is a homogenizing force, for example, possibly because it leads to knowledge sharing about occupational characteristics, such as the amount of education required or average earnings. But as Harding (2011) noted, there are also numerous counterforces that could diversify opinions. In the case of occupations, we suspect that heterogeneity in beliefs within a group represents both differences in substantive opinions about occupational status and proportionally more members who pile-sort somewhat randomly because they view occupations as relatively unimportant status markers.
Moving forward, this decomposition of consensus approach could easily be applied to status beliefs, logics of worth, and cultural frames in any topic space. When doing so, we urge researchers to consider whether there is a general trend of greater belief consensus among those in structurally advantaged positions and whether this difference in consensus can help explain different aspects of inequality. First, agreement on status orders can promote social closure or group solidarity, both of which could have implications for a group’s ability to lobby effectively for their interests (see e.g., Weeden 2002). To this point, our results raise the question of whether a small group demonstrating consensus is sufficient to be perceived as what “most people” think vis-à-vis third-order status beliefs (Correll et al. 2017; Ridgeway and Correll 2006) if no other kind of consensus is apparent in the population. Second, those in structurally advantaged positions wield more institutional and organizational power, and if their viewpoints tend to be homogeneous, they may face little barrier to institutionalizing self-protecting rules and practices. One example of this might be the practice of employers making hiring decisions based on an applicant’s degree profile rather than their relevant skill set (Bills 2003; Jenkins 2023; Rivera 2011).
In closing, we note that our key findings about the structure of agreement parallel those from Boutyline and Vaisey’s (2017) study of political positions, which showed that the aggregate organization of culture principally reflects the positions of an advantaged minority, whereas much of the public lacks consistent organization. These results together call for greater attention to the role resources and social positions play in fostering cohesive culture regardless of the content (Martin 2002).
Supplemental Material
sj-docx-1-spq-10.1177_01902725241256378 – Supplemental material for Intersectional Group Agreement on the Occupational Order
Supplemental material, sj-docx-1-spq-10.1177_01902725241256378 for Intersectional Group Agreement on the Occupational Order by Freda B. Lynn, Yongren Shi and Kevin Kiley in Social Psychology Quarterly
Supplemental Material
sj-pptx-2-spq-10.1177_01902725241256378 – Supplemental material for Intersectional Group Agreement on the Occupational Order
Supplemental material, sj-pptx-2-spq-10.1177_01902725241256378 for Intersectional Group Agreement on the Occupational Order by Freda B. Lynn, Yongren Shi and Kevin Kiley in Social Psychology Quarterly
Footnotes
Acknowledgements
We are indebted to Fabien Accominotti, Cecilia Ridgeway, Michael Sauder, Regan Smock, Hannah Zadeh, and participants of the Colloquium on Analytical Sociology at the European University Institute for their insightful comments on this study. We also wish to thank Lauren Valentino for providing us with the 2012 General Social Survey occupational ratings data.
2
This includes agreement on content and also possibly agreement on whether to approach the task as a first-order (what do I think) versus third-order (what do I think most people think) exercise (Ridgeway and Correll 2006; see also 
:52).
3
Brightness can refer to consensus within a group of evaluators, which is the focus of this study. From this point of view, blurriness is produced by perceivers using different classification systems. Brightness can also be defined with respect to an individual’s own level of certainty around how to rank objects. We cannot capture the latter in this study given that we only have the results of the pile-sort task and not deliberation or uncertainty related to the sorting process itself.
4
We are unable to examine more refined intersectional groups due to sample size limitations.
5
Appendix A in the provides a descriptive summary of all groups in both years.
6
See Appendix B in the for details.
7
See Appendix B in the for details.
8
See Appendix C in the for consensus results for race, gender, and education as single-characteristic groups, along with age, region, political party, religion, and social class for further reference.
9
In Appendix D in the , we confirm that no groups exhibit a high degree of consensus because they sorted all occupations into the same rung.
10
Visualizations were created using igraph (Csárdi et al. 2023) with force atlas 2 () as the layout algorithm.
11
Appendix E in the shows that the same pattern appears if using other measures of structural advantage.
12
Group size is based on number of respondents in the full General Social Survey 2012 sample. The substantive results are the same when size is based on the subset of respondents receiving and responding to the occupational perceptions module.
13
Appendix E in the replicates this analysis for the 1989 data, and the results are substantively equivalent.
Bios
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
