15 (1981) 125-150 125 ElsevierScientific PublishingCompany, Amsterdam - Printed in The Netherlands Qualify and Quantity,
THE STRUCTURE
OF OCCUPATIONAL
INEQUALITY
*
WILLIAM T. BIELBY Department
of Sociology,
of California,
University
Santa Barbara,
U.S.A.
ARNE L. KALLEBERG Department
of Sociology,
Indiana
University,
U.S.A.
Inequalities associated with a system of social stratification result from two related but analytically distinguishable social processes; differential rewards associated with different positions in the social system, and the process of allocation of individuals to those positions. The former process concerns positional inequality; the latter process concerns the movements of individuals among positions. Over the past fifteen years, research on social inequality in the United States has focused almost exclusively upon the movements and attainments of individuals in an exogenously given (and often unspecified) occupational structure. Recently, however, a veritable avalanche of research purports to demonstrate how such structural factors as class (Wright and Perrone, 1977; Wright, 1978; Kalleberg and Griffin, 1978), authority (Kluegel, 1978; Robinson and Kelley, 1979; Wolf and Fligstein, 1979), and economic sector (Bibb and Form, 1977; Beck et al., 1978; Hodson, 1978) affect the process of individual attainment. At least implicitly, most of this research accepts the notion that the structural relationships among positions of “empty places” (Przeworski, 1976; Burawoy, 1977) are causally prior to the process of individual attainment and that “an understanding of positional inequality is logically prior to an examination of allocative processes” (Kalleberg and Griffin, 1980). Yet little empirical research focuses on positions per se, perhaps because we have yet to agree on whether the appropriate “empty places” are jobs, * This researchwassupported in part by the Institute for Researchon Poverty at the University of Wisconsin-Madison,under funds granted by the Economic Opportunity Act of 1964 and administeredby the Department of Health, Education, and Welfare. The opinions expressedherein are those of the authors. 0033-5177/81/0000-0000/$02.50
01981
Elsevier
Scientific
Publishing
Company
126 occupations, classes,sectors, or some other unit of analysis. Nearly all existing research at the positional level of analysis examines attributes of the relationships among occupations. Theoretically, they are the building blocks for most non-Marxist conceptualizations of social structure in industrial societies. For example, Blau and Duncan (1967, p. 7) suggest that the “occupational hierarchy” is “a major source of the various aspects of social stratification in industrial society,” providing the basis for prestige strata, economic classes,and the power structure. Some of the earliest empirical research on positional inequality examined the prestige and status rankings among occupations (e.g., Inkeles and Rossi, 1956; Duncan, 196 1; Hodge, Siegel and Rossi, 1964; Hodge, Treiman and Rossi, 1966). More recently, crosssectional research has related occupational prestige to authority, skill complexity, training requirements and to the sex and race compositions of occupations (e.g., Siegel, 197 1; Stolzenberg, 1975; McLaughlin, 1978; Cullen and Novick, 1979; Spaeth, 1979), while a growing body of longitudinal research has attempted to evaluate theses of occupational “upgrading” and “deskilling” by focusing on the association between changes in the occupational distribution of individuals and changes in the skill, race, and/or sex compositions of occupations (e.g., Scoville, 1969; Berg, 1971; Snyder and Hudis, 1976; Dubnoff, 1978; Snyder et al., 1978; Spenner, 1979). In short, many identify occupations as the “empty places” in a conceptualization of the structure of inequality, and this notion has sustained nearly all of the existing empirical research at the positional level of analysis. While occupations are not necessarily the only or the ideal positional unit of analysis, they certainly cannot be ignored in any reasonable account of the “structure” that is presumed to be logically and causally prior to the processof individual socioeconomic attainment. In our account of the structure of occupational inequality, occupations are positions in the technical division of labor. Specifically, an occupation is an aggregation of jobs in which similar tasks are performed and which have similar technical requirements (cf. Reiss, 1961, pp. lo11; Siegel, 1971, pp. 8, 149-15 1). Occupations also carry differential accessto various resources they can bring to bear in the workplace and in the market; consequently, while they are defined in terms of technical relations, occupations are invariably embedded in social relations. Finally, occupations are differentially rewarded. In our simple threedimensional scheme of the structure of occupational inequality, differences in yewards among occupations can be attributed to inequalities in requirements and resources. Occupational rewards can be either extrinsic or intrinsic. Extrinsic
127 rewards are an incumbent’s claim on the value of the goods and services that are outputs of the technical production process. Earnings are obviously the largest component of extrinsic rewards for most occupations. However, attributes such as security of employment and promotion opportunities - which, in effect, provide claims on the value of future outputs - are also extrinsic rewards. Intrinsic rewards are those that derive from the nature of the occupational task itself and do not involve explicit claims on the value of output. These can include opportunities for self-direction, the degree to which work is challenging or interesting, social and physical conditions of work, and social esteem. Occupational requirements are the technical skills and training necessary for incumbents to perform the occupational task competently. These requirements might include years of schooling, specific vocational training required to perform tasks, and ability to handle the degree of complexity in dealing with people, symbols and physical materials demanded by the tasks. A considerable body of neoclassical economic theory (e.g., Mincer, 1970) and functionalist sociological theory (e.g., Davis and Moore, 1945) suggests that inequality in rewards is attributable to inequality in requirements. For example, Davis and Moore (1945, p. 243) assert that unequal intrinsic and extrinsic rewards “unconsciously evolve” to motivate individuals with the requisite talent and training to enter the appropriate positions. In c.ontrast, we assume neither that the inequality in rewards among occupations is completely technical in origin nor that its evolution is completely “unconscious.” Instead, we assume that occupational groups have differential accessto resources, allowing (1) claims upon the value of the output of the production process, and (2) control over the production process that can affect intrinsic rewards (cf. also Form and Huber, 1976, p. 733). Some resources may be specific attributes of occupational groups, for example, some types of organized bargaining power. Others, such as ownership and authority, are not per se attributes of occupations. While incumbents of a given occupation are not perfectly homogeneous with respect to these characteristics, the attributes vary considerably across occupations and greatly affect the ability of occupational groups to command intrinsic and extrinsic rewards. Similarly, some occupational groups - e.g., those located in large bureaucracies - are more likely to have access to organizational resources, although organizational characteristics are not strictly attributes of occupations. While these resources may not belong in a taxonomy of occupational attributes, they have a central role in any account of occupational inequality. To summarize, an analysis of dimensions of inequality among occu-
128 pational positions is one way to construct a notion of the “structure” that is presumed to underlie the process of individual attainment. At present, the conceptualization of such a structure constitutes a lacuna in the stratification literature. We suggest that the structure of occupational inequality can be modelled by exploring the manner in which differences in occupational rewards - both extrinsic and intrinsic - are generated by inequalities among occupations in requirements and resources. Below we present a preliminary cross-sectional model of occupational structure. It is offered not as a conclusive account of the elusive “structure” of inequality but as a demonstration of the viability of quantitative empirical research at the positional level of analysis. We hope to demonstrate that researchers can do more than assessthe coltsequences of structure for individual attainment; they can - and should - examine the structure itself.
Data and Variables
Our analysis is based on data from the 1972-73 Qudity of EmploySurvey conducted by the Institute for Survey Research at the University of Michigan. These data have been used in several assessments of the consequences of “structure” for individual attainment (e.g., Wright and Perrone, 1977; Bibb and Form, 1977; Stolzenberg, 1978; Kalleberg and Griffin, 1978) and in aggregate form provide a unique opportunity to examine the structure of occupational positions. Attributes of forty occupational groups (see Table I) were constructed from data on 1485 employed individuals concerning the social and economic conditions of their employment. Because this is a first attempt at empirical research on occupational inequality, we devote considerable attention to simple descriptive elaboration of our measures of occupational rewards, requirements, and resources before examining the interrelationships among those dimensions. While the Qudity of Employment Survey is one of the few sources of data containing information on occupational resources as well as rewards, it does impose several limitations upon our analysis. It requires that we aggregate individual responses to obtain “structural” characteristics of occupational groups. This aggregation need not invalidate our measures as properties of social collectives (cf. Lazarsfeld and Menzel, 1969). Indeed, some characteristics of a social structure - for example, the unemployment rate - are, inherently, aggregated characteristics of individuals. Nevertheless, particularly with respect to occupational resources, superior measures could be obtained directly from informament
129 tion on occupations, industries, and firms instead of from a social survey of individuals. A second limitation is that the size of the Quality of EmpZoyment Suruey sample, and the Census classification system, imposed some
TABLE I Occupational Groups Groups
N
1. Protective service workers 2. Personal service workers - retail trade 3. Personal service workers - other 4. Health service workers 5. Laborers (excluding farm) 6. Farmers 7. Transport equipment operatives (except truck drivers) 8. Truck drivers 9. Operatives - manufacture of durables IO. Operatives - manufacture of nondurables II. Operatives - other 12. Craft workers - supervisors, n.e.c. 13. Craft workers - construction 14. Craft workers - manufacture of durables 15. Craft workers - manufacture of nondurables 16. Craft workers - transportation, communication, public utilities 17. Craft workers - other 18. Clerical workers - secretaries 19. Clerical workers - manufacturing 20. Clerical workers - transportation, communication, public utilities 21. Clerical workers - wholesale and retail trade 22. Clerical workers - finance, insurance, real estate 23. Clerical workers - public administration 24. Clerical workers - other 25. Sales - finance, insurance, real estate 26. Sales - wholesale trade, other 27. Sales - retail trade 28. Managers and administrators - construction 29. Managers and administrators - manufacturing 30. Managers and administrators - wholesale trade 31. Managers and administrators - retail trade 32. Managers and administrators - finance, insurance and real estate 33. Managers and administrators - public administration 34. Managers and administrators - other 35. Professional and technical workers - engineers 36. Professional and technical workers - higher education and science 37. Professional and technical workers - health professionals 38. Professional and technical workers - other education 39. Professional and technical workers - technicians, various 40. Professional and technical workers - other
28 36 II 24 53 37 28 28 92 71 37 34 58 44 23 20 47 55 25 25 20 28 22 43 23 13 36 20 25 23 71 17 21 55 39 20 30 65 21 51
130
practical limitations upon our definitions of occupational groups. Conceptually, occupations represent aggregations of jobs that involve similar tasks. If occupational groups are meaningful social actors, then there could be some appropriate level of aggregation at which those groups are reasonably homogeneous with respect to task requirements, rewards, and resources. Unfortunately, even the detailed Census categories for occupation and industry were not constructed to capture homogeneity on any of these three dimensions (cf. Siegel, 1971, pp. 153-l 74). The problem of heterogeneity in the Census classification is compounded by our decision to aggregate even further in order to maintain a minimum of 20 sample casesin each occupational group. The 40 occupational groups listed in Table I were constructed primarily from the Census major occupation and industry group-aggregation of the detailed three-digit classifications. When the number of sample cases permitted, more detailed information was used to make substantive distinctions among occupations. (For example, the category “truck drivers” represents a single three-digit occupational category, and “clerical workers - secretaries” represents six three-digit categories [ I] .> Within the constraints imposed by sample size and the heterogeneity in the classification system, we attempted to preserve the functional task homogeneity of occupational groups while also allowing for differentiation on our three dimensions within occupations of nominally equivalent major group title. For example, among craft workers other than supervisors, we allow for five distinctions by industry. Not only are the task requirements and skills of craft workers likely to vary across these industries, but the resources available to craft workers (e.g., the extent and type of unionization) may differ as well. Consequently, as social actors these five categories of craft workers may function as distinct occupational groups with differential rewards attributable to their different configurations of requirements and resources. For similar reasons distinctions were made within many other major occupational categories. Requirements
For each of the forty occupational groups, we have three measures of occupational requirements. The mean educational attainment (ED) of individuals in the occupational groups indicates the certification requirements of the occupation as well as any required cognitive and noncognitive characteristics that may be indexed by educational attainment. The mean Specific Vocational Preparation score (SVP) is constructed from United States Department of Labor (1968) assessments
131 of the training time required to perform adequately the tasks associated with a job. Only training specifically related to vocational requirements is included in these assessments.The mean General Educational Development score (GED) is constructed from Department of Labor assessments of the level of reasoning with respect to dealing with people, data, and things required to perform adequately in a job [ 21. The three measures of occupational requirements are expressed in standard formdeviations from their respective means in standard-deviation units - for all analyses reported here. The relatively large correlations among the three indicators of occupational requirements (0.56-0.86; see Table III) suggest that all three may to some degree be measures of a single overall dimension of task requirements. However, we also expect that there exist components unique to each of the three measures; general requirements, vocationally specific training and educational preparation and certification are certainly distinct conceptually. The GED and SVP scores might also differ from mean educational level since the former are direct assessments of occupational requirements while the latter is not. (Of course, this problem is negligible if the discrepancy between the mean education of incumbants and the required educational level is nearly constant across occupational groups; see Siegel, 1971, pp. 267-270.) Furthermore, GED and SVP scores may be biased by raters’ perceptions of the educational levels of occupational incumbents. Consequently, for both analytical and methodological reasons it may be useful to examine the relationship between education and the Department of Labor ratings and the residual variation in those ratings that is not associated with mean education. Figure 1 presents a plot of specific training versus education and Fig. 2 a plot of general requirements versus education for the 40 occupational groups. All managerial and professional groups (28-40) are located above the mean on both GED and SVP (nearly all are at least one-half standard deviation above the mean). These thirteen groups are somewhat more differentiated with respect to education, ranging from just below the mean (managers in construction (28) and wholesale trade (30)) to about two and one-half standard deviations above the mean (professionals in education (36,38)). Occupations located considerably below the mean on all three measures are personal service, laborer, and operative groups (2, 3, 5, 7-11). Health service workers (4), clerical workers in manufacturing (19) and trade (21), and retail salesworkers (27) - groups that might be considered part of a “white collar” working class - also fall below the mean on all three requirement measures. The remaining clerical groups, protective service workers (I) and whole-
132
316 (N=40)
-hi-
, (2.0)
I (2.5)
I (3.0) ED
Fig. 1. Plot of Occupational Inequality in Education (ED) and Specific Vocational Preparation (SVP).
sale sales workers (26), are located higher with respect to education than they are with respect to specific training or general requirements, suggesting that perhaps for these groups educational certification serves more as a screening device than an indicator of either specific or general occupational skills. The solid diagonaI line in Fig. 1 represents the linear regression of SVP scores on mean group education. Note that farmers (6) and craft groups (12-17) show quite large residuals. These groups require substantially more specific vocational training than would be predicted on the basis of mean educational level. Indeed, when these seven groups and the two groups with unusually high education levels (professionals in education (36,38)) are ignored, the proportion of variance in SVP scores attributable to educational level nearly doubles from 0.32 to 0.63 (see the dotted diagonal line in Fig. 1). (A similar but much smaller effect occurs in the relationship between GED and education when the same groups are omitted; see the two diagonal lines in Fig. 2.) Except for farming and crafts groups, occupational groups appear to be
133 4 GED 42.0)
I (3.0)
* ED
d-1.5)
.5
Fig. 2. Plot of Occupational Inequality in Education (ED) and General Educational Development (GED).
distributed along a continuum from low to high in specific, general and educational requirements. With respect to occupational requirements, the “disadvantaged” jobs are associated with service, skilled, and semiskilled manual occupations, and the “new working class” nonmanual occupations - clerical and retail sales. The “advantaged” nonmanual jobs are managerial and professional occupations. Farming and craft occupations hold a unique position in the contemporary occupational structure, requiring specific vocational training but not educational certification [ 31. Indeed, given their anomalous configuration of technical requirements, it is not surprising that jobs in these occupations are among those disappearing most rapidly in “postindustrial” society. While the three measures of requirements are all to a degree empirically as well as conceptually distinct, the relatively high intercorrelations among them and the relatively small number of groups preclude assessment of their individual effects on occupational rewards. Consequently, we combined them into a single, composite requirement scale
134 (REQ), weighting the three component measures by a canonical correlation analysis in order to maximize the linear association between REQ and our two reward measures (INTR and EXTR; see below). The standardized canonical weights were 0.085 for ED, 0.227 for GED and 0.740 for SVP; the canonical correlation was 0.819. (Nevertheless, the individual measures should be included in any work utilizing less-aggregated units of analysis, and multiple-indicator models of occupational requirements examined for “contamination” of the indicators by one another or by an overall “status” dimension.) Resources Three measures of occupational resources are included in our analysis. The authority position of an occupational group (SUP) is measured by the proportion of occupational incumbents who supervise the work of others. The bargaining strength of an occupational group (UNION) is measured by the proportion of occupational incumbents belonging to a union or employees’ association. The average size of the organization in which occupational incumbents work (SIZE) in used as a proxy for organizational resources available to an occupational group. Our measures of authority position, bargaining strength, and organizational resources are far from ideal. In future work, measures of authority position should include indicators of decision-making power in hiring and firing and in pay and promotion. The bargaining strategy of unions depends on the structure of the labor market; for example, considerable differences often exist between the strategies of craft unions and industrial unions (Doeringer and Piore, 1971, pp. 148150). Employees’ associations for professional occupations often rely more upon restrictions on entry to the occupation through licensing than do labor unions. Future .conceptual elaboration and empirical research will need to incorporate these different sources of organized labor market power. While scale of operations as measured by number of employees is perhaps the best single easily obtained indicator of the sectoral location of an organization (Averitt, 1968), a more complete view of organizational resources would include direct measures of capital intensity, automation of production processes,long-run organizational growth and profit strategies, and the “market position” of the organization with- respect to its (material, human, and informational) inputs and outputs. Furthermore, not all groups have equal accessto organizational resources, and the organizational resources themselves may differ depending upon the technical production process and the hierarchical authority structure.
135 Plots of supervisory position by size, and union bargaining strength by size appear in Figs. 3 and 4 (all measures expressed in standard form). Relatively “powerless” groups, those at least one-half standard deviation below the mean on both union and supervisory resources, are personal service occupations (2, 3), secretaries and clerical workers in finance, insurance, or real estate (18, 22), and sales workers not in finance, insurance, or real estate (26, 27). Occupational groups at least one-half standard deviation above the mean on supervisory position include craft supervisors (I 6), all managerial groups (28-34), and all professional groups except technicians (39) and educators not in higher education (38; mostly primary and secondary school teachers). The relatively unionized or otherwise organized occupational groups are protective service workers (I), laborers other than farm laborers (5), operatives other than the residual category (7-1 U), craft workers other than supervisors (13-l 6), clerical workersin manufacturing, in transportation, communications and public utilities, and in public administra-
A SUP
-(2L3) 52 2-8 2-s -(1.5) 3;
3-2 3-O
J
YO -(I*))
A
25
j7 3;
b
-0.5) i
5 I (-2.5)
I (-2.0)
I C-1.5)
I
i
I (-0.5)
(-1.0)
I (0.5)
I
I 95)
(1.0)
i7
I;
ir
2-S
I (20)
l-5
m 21 k6.5) 7.8 * 27
;4
. ju
I (2.5) SIZE
39
i Ib
4-1.0)
5
i
.i 23
i 17.
i0 j,
a
-C-1.5)
Fig. 3. Plot of Occupational Inequality in Supervisory Position (SUP) and Organizational Size (SIZE).
-
136 UNION i6 -(2D)
is
-(lx))., rb
i3
is
is
53 -co.51
I
I
I
I
(-2.5)
(-2.0)
(-1.5)
(4.0)
I 95)
I (0.5)
12 24
li
I (1.0)
I (1.5)
I (2.0)
$7 io 2;
t 3:
.2j
I (2.5) SIZE
-(-a5)
>A . -(-1.0)
.%
35
ia 4-1.5)
Fig. 4. Plot of Occupational Inequality in Union Bargaining Strength (UNION) Organizational Size (SIZE).
and
tion (19, 20, 23), and educators not in higher education (38). Occupational groups that tend to be located in large organizations (again, about one-half standard deviation or more above the mean) are operative, craft, clerical and managerial groups involved in manufacturing (9, 10, 14, 1.5, 19, 29), health service workers (4), engineers and technicians (35, 39), and clerical workers in transportation, communications and public utilities, and in public administration (20, 23). Rewards We were more successful in achieving a correspondence between concept and measurement for occupational rewards. Our measure of intrinsic rewards is the group mean on a scale composed of six items concerning rewards that relate to the nature of the occupational task whether the work is interesting, challenging, allows for autonomy, etc.
-
137 (see Kalleberg, 1977, for a discussion of the intrinsic-rewards scale). Our measure of extrinsic rewards is the group mean on a canonically weighted composite of total annual income [4] and a three-item scale assessingopportunities for promotion. A canonical correlation analysis provided weights for the income and promotion measures so that the composite would have maximum linear association with the three measures of occupational resources and the three measures of occupational requirements. The standardized weights were 0.789 for the income measure and 0.300 for the promotion measure; the canonical correlation was 0.772. Figure 5 is a plot of intrinsic and extrinsic rewards for the 40 occupational groups (both variables measured in standard form). Only 6 of the 40 groups are above the mean on one reward and below the mean on the other, and none is at least one-half standard deviation below the mean on one and at least one-half standard deviation above on the other (the correlation between the two rewards is 0.774; see Table III). All the managerial groups and all the professional groups except that \
\
INTR y’hd
.
QA \ function
‘\\
t
(1.5)
\ \
I (-2.0)
I (-1.5) 3
cl.01
ie ‘\
9-K
\ I (-2.5)
.54 26
i0 \
I (-1.01 . 27
Fig. 5. Plot of Occupational Rewards.
2.2 p . (-0.2 5) ’
\ .C
L?,i /\ \,
,
37
./I\ ‘.
/ /’
Inequality
1
\
\
I (0.5)
/L&t 2-5 /’
6
* ‘2
’ ,y;9
58 :I?
I (1.0)
/
/
/
DA function
/
I (1.51
I (2D)
I (2.5.5)) EXTR
j9 \
in Intrinsic (INkR)
and Extrinsic (EXTR)
I38 composed mostly of primary and secondary school teachers (38) are above the mean on both rewards (although managers in finance, insurance, and real estate (3.2) and health professionals [5] (37) are only slightly above the mean on each reward), Also above the mean on both rewards are the two groups with the highest level of intrinsic rewards: farmers (6), and salespersons in finance, insurance, and real estate (25). The latter group also has the highest level of extrinsic rewards. Finally, salespersons in wholesale trade (26), craft supervisors (12), and craft workers in construction (13), and in transportation, communication, and utilities (16) are above the mean on both rewards. Groups considerably below the mean on both rewards encompass every major occupational group except craft workers, managers, and professionals. For example, personal service workers in retail trade (2), health service workers (4), laborers other than farm laborers (.5), operatives in manufacturing and other operatives (9-11), secretaries, clerical workers in retail trade, and other clerical workers (18, 21, 24), are all at least one standard deviation below the mean on one reward and one-half standard deviation below the mean on the other. Poorly rewarded jobs subsume many kinds of occupational tasks in the contemporary occupational structure. Because both measures of rewards are averages of the individual responses of occupational incumbents, we performed several analyses to determine the degree to which systematic inequality in the measures occurs between rather than within our 40 occupational groups. First, we computed that 23% of the variance in intrinsic rewards (INTR) and 26% of the variance in extrinsic rewards (EXTR) occurs among the 40 occupational groups. While most variation occurs within the 40 groups, not all of it reflects systematic individual-level variation in occupational rewards attributable to individual-level response and requirement determinants. Some of it is attributable to occupational-level variation that occurs within our heterogeneous aggregate categories, and some is due to individual response error in measuring the rewards (see Kalleberg, 1975, for a discussion of the reliability of the measures used here). A regression analysis of the within-group covariation revealed that only 5% of the within-group variance in intrinsic rewards and only 8% of the within-group variance in extrinsic rewards was accounted for by withingroup variation in requirements and resources. Thus, while considerable variation in rewards occurs within our 40 occupational categories, most of the systematic variation in the rewards seems to be captured by our 40 groups. In addition to the univariate analyses of the reward measures, we also computed a multivariate discriminant analysis of the two measures in
139 TABLE II Discriminant Analysis Measures of Occupational Group Differences in Rewards Standardized Discriminant Coefficients a
INTR EXTR Canonical p % Discrimination
DA Function 1
DA Function 2
0.465 0.667 0.544 76.9
-0.567 0.495 0.355 23.1
a Coefficients are standardized relative to withingroup variation. Coefficients standardized with respect to between-group variation are 0.426 and 0.720 for the first function, and -0.520 and 0.534 for the second function. Both functions are statistically significant at the 0.001 level.
order to examine the underlying dimensions of between-group variation in occupational rewards. Discriminant analysis selects successive orthogonal linear composites of intrinsic and extrinsic rewards that have maximum variation between the 40 groups relative to their variation within the 40 groups (see Tatsuoka, 197 1, pp. 157-l 83). The results of the analysis are reported in Table II. Both discriminant functions were statistically significant, although more than three-fourths of the total discriminatory power is attributable to the first function. The two functions are indicated by the dotted diagonal lines in Fig. 5. (The location of each group on a discriminant function can be obtained by projecting its location perpendicularly to the dotted line.) The first function, that differentiates maximally among the groups, discriminates groups having more of both rewards from those having less of both rewards, and it weights extrinsic rewards slightly more heavily than intrinsic rewards. Sales workers in finance, insurance, and real estate (25) score highest on this function, service workers in retail trade (2) and operatives in manufacture of nondurables (10) score lowest. The second discriminant function differentiates groups with respect to the “trade-off” between intrinsic and extrinsic rewards, weighting the (standardized) intrinsic and extrinsic reward measures almost equally but with opposite sign. Clerical workers in public administration (23) are located at the extrinsic extreme of the trade-off, while miscellaneous personal service workers (3), teachers not in higher education (38), and farmers (6) are located almost equally at the intrinsic extreme. To summarize, the discriminant analysis provides a statistical rationale - between-group relative to within-group inequality - for the above descriptive analysis of the location of the groups in the two-dimensional plot of rewards. It suggests that the 40 groups are differentiated in their rewards along two
III
1. REQ 2. ED 3. GED 4.svP 5. SIZE 6. SUP 7. UNION 8. INTR 9.EXTR 10. PRSTG
Correlations
TABLE
,
Among
0.691 0.938 0.984 -0.098 0.834 -0.288 0.791 0.746 0.869
1
Measures
0.834 0.563 0.088 0.556 -0.216 0.541 0.525 0.860
-
2
of Occupational
-0.135 0.827 -0.256 0.782 0.729 0.791
0.735 0.713 0.923
4
Rewards,
0.865 -0.024 0.768 -0.354
-
3
Requirements,
-0.501 -0.132 0.112
-0.286 0.506
5
and Resources
0.703 0.694 0.757
-0.445
6
(N = 40)
-0.382 -0.190 -0.245
1
0.774 0.601
8
0.651
9
-
10
141 dimensions, primarily with respect to the “goodness” or “badness” of both rewards, but also with respect to the “trade-off” between intrinsic and extrinsic rewards. Having described the location of our 40 occupational groups with respect to occupational requirements, rewards, and resources, we now present several models that assessthe manner in which inequality in occupational rewards can be attributed to inequality in requirements and resources. All models are estimated from the correlations presented in Table III. First we specify that inequality in intrinsic and extrinsic occupational rewards is attributable to inequality in resources and requirements. In a second model we argue that organized bargaining strength should not be considered a determinant of intrinsic rewards and should perhaps be considered an outcome of those aspects of occupational tasks that determine intrinsic rewards. A third model assumes that occupational prestige scores measure a third type of occupational reward and specifies prestige ratings to be determined by occupational requirements and resources. Finally we present a canonical correlation analysis to test whether occupational resources and requirements affect both intrinsic and extrinsic rewards through a single intervening process. Estimates for the first model appear in columns 1 and 3 of Table IV. The model accounts for about four-fifths of the between-group variance in intrinsic rewards (INTR) and about three-fifths of the between-group TABLE IV Standardized Regression Coefficients for Models of the Relationship of Occupational Rewards to their Determinants (N = 40) Independent variables
l.REQ 2. SIZE 3. SUP 4. UNION 5.INTR 6.EXTR R*
Dependent variable a 1 INTR
2 INTR
3 EXTR
0x33* -0.483* -0.149 0.050 0.817
0.888’ -0.463+ -0.169 0.815
0.544* -0.060 0.277 0.122 0.583
4 UNION -0.068 0.481* -0.374 0.176 0.362
5 PRSTG 0.668* 0.294’ 0.242 -0.093 0.819
6 PRSTG 0.600’ 0.337* 0.262 -0.095 0.091 0.023 0.821
a Asterisk indicates a rejection probability of less than 0.01 for the conventional t-test of the hypothesis that a coefficient is zero. The tests are merely suggestive and should be interpreted with caution, since the data are not from a sample of occupations and are aggregated from the responses of occupational incumbents.
142 variance in extrinsic rewards (EXTR). Inequality in rewards related to the nature of occupational tasks (intrinsic rewards) is largely attributable to the requirements of those tasks (REQ) according to our first model. Neither authority position (SUP) nor organized bargaining strength (UNION) appear to have been used by occupational groups to enhance occupational rewards [ 61. Occupational groups that tend to be located in large organizations (SIZE) are likely to obtain fewer intrinsic rewards, net of occupational requirements, union resources, and supervisory resources. While none of the three resources appears to have been utilized by occupational groups to increase intrinsic rewards, the authority position of an occupational group appears to make a modest contribution to extrinsic rewards, and organized bargaining strength has a small effect upon extrinsic rewards. As with intrinsic rewards, occupational requirements make the strongest relative contribution to inequality in extrinsic occupational rewards. However, for several reasons we hesitate to attribute the effects of requirements to “technical-functional” processes and the effects of resources to “conflict-power” processes.The components of the requirement measure are more complete and of better quality than the three measures of occupational resources, and the requirement composite was constructed to have maximum association with the reward measures. Furthermore, the requirement measure subsumes a degree of occupational-group screening and monopolization of occupationally specific skills, two aspects of resources that the model attributes to requirements. Nevertheless, future research should be able to disentangle the distinct influences of requirements and resources, and it should also explicitly model the interaction and reciprocity between requirements and resources in their effects on occupational rewards. While we suspect that organizational variables (e.g., SIZE) interact with requirements and other resources, the small number of occupational groups and the colinearity among measures precludes the presentation of stable assessmentsof such effects here 171. Giddens (1973, pp. 207-215) and others (Parkin, 1971, p. 91;Galbraith, 1973, pp. 107, 175, 289-291; Braverman, 1974, pp. 10, 150) have noted that in advanced capitalism, labor unions become increasWY “economistic,” oriented toward increasing labor’s share of the value of output instead of toward labor control over the production process. The negligible effect of unionization on intrinsic rewards detected in our first model is consistent with this view. If contemporary union strategy relinquishes control over the structure of the technical production process, then perhaps organized bargaining strength is more properly seen as an outcome of working conditions than as an
143 exogenous determinant of intrinsic rewards that are attributable to the structure of work. Estimates for a revised model consistent with this reasoning appear in columns 2, 3, and 4 of Table IV. Omitting UNION from the INTR equation (column 2 of Table IV) hardly changes the other coefficients and requires no modification of our interpretation of the determinants of occupational inequality in intrinsic rewards. In column 4 of Table IV, organized bargaining strength is expressed as a function of intrinsic rewards, occupational requirements, authority position, and organizational size. As expected, groups lacking authority resources and groups that tend to be located in large organizations (e.g., in manufacturing industries) are likely to be more organized. However, this revised model does not support the suggestion that intrinsically unrewarding tasks contribute to unionization or similar organization of an occupation. While occupations with intrinsically unrewarding jobs tend to be more unionized (the zero-order correlation of INTR with UNION is -0.382; see Table II), in the revised model the net effect of intrinsic occupational rewards upon organized bargaining strength is positive (0.176). It may be that an appropriate model for a static crosssectional analysis should allow for reciprocal causation between the intrinsic nature of the task and unionization, but a fully satisfactory representation would require a dynamic model of historical data that incorporates both the antecedents and consequences of occupational organized bargaining strength over time. Columns 5 and 6 of Table IV present estimates for regression equations where the mean Siegel prestige score (Siegel, 1971) for each occupational group is the dependent variable. The first equation could be,. considered part of a model of occupational inequality where the prestige score (PRSTG) is a measure of a third occupational reward in addition to intrinsic and extrinsic rewards. Comparing the estimates in column 5 with those in columns 1 and 3, it can be seen that the determinants of the prestige score parallel neither those of intrinsic rewards nor extrinsic rewards. As with the latter two measures, occupational requirements have the largest relative effect on the mean prestige score, but the organizational size context of an occupational group has a modest positive effect. Authority position has a small positive effect upon the mean prestige score, similar to its relationship to extrinsic rewards but unlike its contribution to intrinsic rewards. Although the empirical evidence is far from conclusive, it may be that the “hybrid” nature of determinants of the mean prestige score compared to those of the other measures of rewards reflects the conceptual ambiguity of prestige ratings as a measure of occupational rewards at the positional level of analysis.
144 The estimates in the last column of Table IV are presented simply as a descriptive assessmentof how mean prestige ratings are predicted from the measures of rewards, requirements, and resources. The requirement measure is the best single predictor of the mean occupational prestige rating, and both organizational size and authority position have modest net effects. The size effect may be an artifact of differential detailed occupational composition across those major occupational groups that we have subdivided by major industry (for example, the detailed occupational composition for managers in retail trade probably differs from that for managers in manufacturing, and the latter group scores higher on both mean prestige rating and mean organizational size). However, this argument should also apply to the intrinsic and extrinsic reward equations, yet no similar positive effects of size were detected in those equations. Another puzzling finding is the lack of a net contribution of either of the reward measures to the mean prestige score, despite the moderate zero-order correlations of each reward measure with the mean prestige rating (0.601 and 0.65 1 respectively for intrinsic and extrinsic rewards; see Table III). Again, the ambiguity of the conceptual meaning of prestige ratings at the occupational level of analysis and the colinearity of the measures make this result difficult to interpret. If prestige ratings can be viewed as synthetic composites of popular perceptions of the goodness of occupations with respect to their requirements, rewards and resources, then rewards should have nontrivial positive net effects in the prediction equation. In a final analysis we investigated whether a single intervening process underlies the manner in which inequality in occupational requirements and resources determines inequality in occupational rewards. A canonical correlation analysis was performed that related the composite requirement measure (REQ) and the three resource measures (SIZE, SUP, UNION) to the two measures of occupational rewards (INTR, EXTR). Should a single intervening process prevail, for example, if the relationships are mediated by an unobservable variable such as “status,” “prestige,” or overall “goodness” of an occupation, then we should detect a single pair of canonical variates relating a linear composite of rewards and resources to a weighted sum of intrinsic and extrinsic rewards (see Hauser and Goldberger, 1971, pp. 106-l 14). If a second nontrivial pair of canonical variates is extracted, the notion of a single intervening process is less tenable. The results presented in Table V show that we detected two meaningful pairs of canonical variates. The first pair of variates weights intrinsic rewards ten times more strongly than extrinsic rewards; that is, the first pair of canonical variates essentially accounts for varation in intrinsic occupational
145 TABLE V Canonical Correlation Representation of the Relationship Between Occupational Rewards and their Determinants; Standardized Canonical Coefficients (N = 40) First canonical variate First Set REQ SIZE SUP UNION Second Set INTR EXTR Canonical p
Second canonical variate
0.991 -0.572 -0.215 0.045
-0.215 0.812 1.074 0.233
1.087 -0.116 0.905
-1.146 1.575 0.565
rewards. Consequently, the standardized canonical coefficients are nearly identical to the regression coefficients reported in column 1 of Table IV (and the first canonical correlation is nearly identical to the corresponding multiple-correlation coefficient), and the interpretation is identical to that for the regression analysis. The second pair of canonical variates weights the two measures of rewards more equally but with opposite signs. Therefore, the second pair of variates might be interpreted as accounting for the “trade-off” dimension of extrinsic versus intrinsic rewards discussed above in the context of the discriminant analysis of occupational rewards. Thus, both authority position (SUP) and organizational resources (SIZE) can be seen as contributing to the extrinsic side of the trade-off, and organized bargaining strength (UNION) appears to have a similar but smaller effect. Occupational requirements (REQ), in contrast, make a small relative contribution in the intrinsic direction. The various statistical models presented above suggest that occupational task requirements are strongly related to both intrinsic and extrinsic occupational rewards (especially the former), and that occupational resources are utilized to make claims upon the value of the output of the production process but are not utilized to manipulate that process to enhance intrinsic rewards. Because of the preliminary nature of the analysis and limitations of the data - the aggregation of occupational groups, the incompleteness of the resource measures, the combining of conceptually distinct requirement measures - these results are less than conclusive. Nevertheless the analysis does provide a preliminary map of the patterns of occupational inequality in requirements, resources, and rewards, and we hope that it will stimulate others to
146 undertake empirical research at the positional level of analysis. By elaborating the conceptualization of the three dimensions of occupational inequality and using more-complete data, researchers should be able to construct a more accurate representation of the structure among the “empty places” in the socioeconomic hierarchy. Conclusions
By examining the interrelationships among the requirements, resources, and rewards associated with occupational positions, we have attempted to construct a preliminary account of the “structure” underlying the process of individual attainment. Using data on 40 occupational groups aggregated from the 1972-73 Quality of Employment Survey, we first described the patterns of inequality for each of the three dimensions and then examined the interrelationships among. the dimensions. A discriminant analysis revealed that occupational rewards are differentiated along two dimensions. The first dimension differentiates occupations highly rewarded both intrinsically and extrinsically from those having little of either type of reward. The second differentiates occupations according to the “trade-offs” between intrinsic and extrinsic rewards. Assessing the determinants of inequality of occupational rewards, we found technical requirements to be strongly related to both intrinsic and extrinsic rewards but with a larger impact on the former. In contrast, occupational groups appear to use resources of authority and organized bargaining strength to enhance extrinsic rewards, but not rewards derived intrinsically from the nature of the occupational task. The patterns of determinants of intrinsic and extrinsic rewards are quite different and cannot be represented by a process with a single unobservable mediating construct. Our analysis has no doubt raised more questions than it has answered. Data limitations precluded exploration of some important issues. For example, it is probably impossible to disentangle the contributions of “technical relations” and “social relations” to reward inequality without longitudinal data. The emergence of such occupational resources as unionized bargaining strength is probably an outcome of past reward (or deprivation) configurations, and the implementation of technical arrangements such as automation or “deskilling” may in fact be an exercise of occupational resources by management (Braverman, 1974). In short, a complete account of the interplay between positional requirements, rewards, and resources requires historical data and dynamic modelling.
147
Finally, a more fundamental issue concerns the appropriate positional unit of analysis. While there is a long tradition of theory and research in the study of social inequality that focuses on occupations, jobs are the units by which work is organized in industrial societies. These jobs are structural within organizations, and much recent research suggests that the institutional, industrial, or market sectors within which those organizations are located have much to do with the structure of social inequality (cf. Baron and Bielby, 1980; Kalleberg and S$rensen, 1979). By focusing on the consequences of “structure” for individual attainment rather than on the structure among positions per se, recent research has avoided facing the theoretical issues of an appropriate positional unit of analysis and a complete specification of the structure that so many of us maintain is logically and causally prior to the process of attainment. We hope the research reported here contributes to redirecting empirical research and theoretical debate, Notes 1 A complete mapping of the three-digit occupation and industry codes into our 40 occupational groups is available upon request. A forty-first group, “farm laborers,” was omitted from the analyses reported here, because it contained only seven sample cases. In two instances, occupational groups were constructed even though they included less than 20 cases. Since occupations of retail and wholesale sales are quite different, we decided to retain the distinction, even though it resulted in only 13 cases in the latter category. For similar reasons we retained the category of “managers and administrators - finance, insurance and real estate” which contains 17 cases. 2 Department of Labor Employment Service personnel provided GED and SVP . ratings for 4000 detailed job titles. These titles were subsequently mapped into Census three-digit occupation categories, and unweighted averages of the GED and SVP scores were assigned to the three-digit occupation codes. We aggregated these scores to the level of our 40 occupational groups, where the average for each group was weighted by the distribution of three-digit occupation codes for that group in the Quality of Employment Survey sample.For further information about the construction and use of GED and SVP scores,seeU.S. Department of Labor ( 1968, pp. 65 l-653) and Horowitz and Herrnstadt ( 1966, pp. 232-240). 3 Figures 1 and 2 suggestthat what Braverman(1974) hasdescribedasa deskilling of the American labor force may in fact be a polarization of skills. There has been a continued increasesince 1900 in the proportion of the nonagricultural labor force in nonmanual occupations with considerablegeneral, specific and educational requirements(professionalgroups, and to a lesserextent, managerial groups), but a more dramatic increasehas occurred (from about 5% of the nonagricultural labor force in 1900 to about 16%in 1960) in the clerical nonmanual occupations that have relatively lower requirements.While the proportion of un-
148
4
5 6 7
skilled laborers has decreased steadily from about 20% of the nonagricultural labor force to about 6% in 1960, the proportion of skilled craft workers appears to have decreased slightly in recent years, and the proportion of unskilled service workers (other than private household workers) has increased from less than 6% to 10% in 1960. The proportion of “semiskilled” operators in the nonagricultural labor force has remained essentially constant at about 20% since 1900 (U.S. Bureau of the Census, 1960, p. 74; 1969, p. 222). Total annual income includes all sources of income, not just occupationally specific earnings. Ideally, we would have prefered a measure of occupationally specific earnings adjusted for hours and weeks worked. The category “health professionals” is composed primarily of technicians, not physicians. This ignores a relational aspect of supervisory resources - they may affect the intrinsic rewards available to other occupational groups. Experimentation with models including three multiplicative interaction terms, SIZE x REQ, SIZE x SUP, and SIZE x UNION, yielded results extremely sensitive to which of the interaction terms were included in the equations. While the magnitudes of the effects were quite unstable, we did consistently detect some interactions. The effect of requirements on intrinsic rewards appears to be larger among those occupational groups located in large organizations, but the effect of requirements on extrinsic rewards appears to be less among those groups. The effect of both authority position and union bargaining strength also appears to be larger among groups located in large organizations. The latter findings, if replicable, suggest that occupational groups can exploit organizational resources through authority position and organized bargaining strength.
References Averitt, Robert T. (1968). The Dual Economy: The Dynamics of American Industrial Structure. New York: W.W. Norton. Baron, JamesN. and Bielby, William T. (1980). “Bringing the firms back in: stratification, segmentation and the organization of work,” American Sociological Review 45: 137-765.
Beck, E.M., Horan, Patrick M. andTolbert, CharlesM., II (1978). “Stratification in a dual economy: a sectoralmodel of earningsdetermination,” American Sociological Review 43: 704-720.
Berg, Ivan (1971). Education and Jobs: the Great Training Robbery. New York: Praeger. Bibb, Robert and Form, William H. (1977). “The effects of industrial, occupational, and sex stratification on wagesin blue-collar markets,” Social Forces 55: 974-996.
Blau, Peter M. and Duncan, Otis Dudley (1967). The American Occupational Structure. New York: Wiley. Braverman, Harry (1974). Labor and Monopoly Capital. New York: Monthly Review Press. Burawoy, Michael (1977). “Social structure, homogenization, and ‘The processof status attainment in the United States and Great Britain’,” American Journal of Sociology 82: 1031-1042.
149 Cullen, John B. and Novick, Shelley M. (1979). “The Davis-Moore theory of stratification: a further examination and extension,” American Journal of Sociology 4: 1424-1437. Davis, Kingsley and Moore, Wilbert E. (1945). “Some principles of stratification,” American Sociological Review 10: 242-249. Doeringer, Peter B. and Piore, Michael J. (197 1). In ternal Labor Markets and Manpower Analysis. Lexington, MA: D.C. Heath. Dubnoff, Steven (1978). “Inter-occupational shifts and changesin the quality of work in the American economy, 1900-1970.” Paper presentedto the Society for the Study of SocialProblems,SanFrancisco. Duncan, Otis Dudley (1961). “A socioeconomic index for all occupations,” pp. 108-138 in Albert J. Reiss,Jr., ed., Occupations and Social Status. New York: Free Press. Form, William H. and Huber, Joan A. ( 1976). “Occupational power,” pp. 15I-806 in R. Dubin, ed., Handbook of Work, Organization and Society. Chicago: Rand McNally. Galbraith, John Kenneth (1973). Economics and the Public Purpose. Boston: Houghton Mifflin. Giddens, Anthony (1973). The Class Structure of Advanced Societies. New York: Harper and Row. Hauser,Robert M. and Goldberger, Arthur S. (1971). “The treatment of unobservable variablesin path analysis,” pp. 8 1- 117 in Herbert L. Costner, ed., Sociological Methodology 1971. SanFrancisco: Jossey-Bass. Hodge, Robert W., Siegel, Paul M. and Rossi, Peter H. (1964). “Occupational prestigein the United States,” American Journal of Sociology 70: 286-302. Hodge, Robert W., Treiman, Donald and Rossi,Peter H. (1966). “A comparative study of occupational prestige,” pp. 309-321 in R. Bendix and SM. Lipset, eds., Class, Status, and Power (2nd edn.). New York: Free Press. Hodson, Randy D. (1978). “Labor in the monopoly, competitive, and state sectors of production,” Politics and Society 8: 429-480. Horowitz, Morris A. and Herrnstadt, Irwing L. ( 1966). “Changesin the skill requirements of occupations in selectedindustries,” pp. 227-287 in The Employment Impact of Technological Change, Appendix Vol. II, Report of the National Commission on Technology, Automation, and Economic Progress. Washington,
DC: U.S. Government Printing Office. Inkeles, A. and Rossi,Peter H. (1956). “National comparisonsof occupational prestige,” American Journal of Sociology 66: l-3 1. Kalleberg, Arne L. (1975). “Work values,job rewards,and job satisfaction: a theory of the quality of work experience.” Unpublished Ph.D. dissertation. University of Wisconsin. KalIeberg, Ame L. (1977). “Work values and job rewards: a theory of job satisfaction,” American Sociological Review 42: 124-143. Kalleberg, Arne L. and Griffin, Larry J. (1978). “Positional sourcesof inequality in job satisfaction,” Sociology of Work and Occupations 5: 37 l-401. Kalleberg, Arne L. and Griffin, Larry J. (1980). “Class,occupation, and inequality in job rewards,” American Journal of Sociology 85: 73 l-768. Kalleberg, Arne L. and S$rensen,Aage B. (1979). “The sociology of labor markets,” pp. 35 l-379 in A. Inkeles, ed., Annual Review of Sociology, Vol. 5. Palo Alto: Annual Reviews,Inc.
150 Kluegel, J.R. (1978). “The causes and costs of racial exclusion from job authority,” American
Sociological
Review 43: 285-301.
Lazarsfeld, Paul F. and Menzel, Herbert (1969). “On the relation between individual and collective properties,” pp. 499-516 in A. Etzioni, ed., A Sociological Reader on Complex Organizations. New York: Holt, Rinehart and Winston. McLaughlin, Steven D. (1978). “Occupational sex identification and the assessment of male and female earnings inequality,” American Sociological Review 43: 909-921. Mincer, Jacob (1970). “The distribution of labor incomes: a survey,” Journal of Economic
Literature
8: l-26.
Parkin, Frank (197 1). Class Inequality and Political Order. New York: Praeger. Przeworski, A. (1976). “The process of class formation: from Karl Kautsky’s ‘Class Struggle’ to recent controversies.” Mimeo. Department of Political Science,University of Chicago. Reiss,Albert J., Jr. (196 1). Occupations and Social Status. New York: Free Press. Robinson, Robert V. and Kelley, Jonathan (1979). “Class as conceived by Marx and Dahrendorf: effects on income inequality, classconsciousness, and classconflict in the United States and Great Britain,” American Sociological Review 44: 38-58. Scoville, J.G. (1969). The Job Content of the U.S. Economy, 1940-1970. New York: McGraw-Hill. Siegel,Paul (197 1). “The American occupational prestigestructure.” Unpublished Ph.D. dissertation.University of Chicago. Snyder, David, Hayward, Mark D. and Hudis, Paula M. (1978). “The location of changein the sexual structure of occupations, 1950-1970: insights from labor market segmentationtheory,” American Journal of Sociology 84: 706-717. Snyder, David and Hudis, Paula M. (1976). “Occupational income and the effects of minority competition and segregation:a reanalysisand somenew evidence,” American Sociological Review 41: 209-234. Spaeth, Joe L. (1979). “Vertical differentiation amongoccupations,” American Sociological
Review
44: 746-762.
Spenner,Kenneth I. ( 1979). “Temporal changesin work content,” American
Socio-
logical Review 44: 968-975.
Stolzenberg, RossM. (1975). “Occupations, labor marketsand the processof wage attainment,” American Sociological Review 40: 645-665. Stolzenberg, ROSS M. (1978). “Bringing the bossback in: employer size, employee schooling, and socioeconomicachievement,” American Sociological Review 43: 813-828. Tatsuoka, Maurice (1971). Multivariate Analysis. New York: Wiley. United States Bureau of the Census(1960). Historical Statistics of the United States, Colonial Times to 19.57. Washington, DC: U.S. Government Printing Office. United States Bureau of the Census(1969). Statistical Abstract of the United States: 1969 (90th edn.). Washington,DC: U.S. Government Printing Office. United States Department of Labor (1968). Dictionary of Occupational Titles (3rd edn.). Volume 2: Occupational Classification. Washington, DC: U.S. Govemment Printing Office. Wolf, Wendy C. and Fligstein, Neil D. (1979). “Sex and authority in the workplace: the causesof sexualinequality,” American Sociological Review 44: 235-252. Wright, Erik 0. (1978). “Race, class,and income inequality,” American Journal of Sociology
83: 1368-1397.
Wright, Erik Olin and Perrone, Luca (1977). “Structural classposition and income inequality,” American Sociological Review 42: 32-55.