The objective was to examine BMI of working-age Canadian adults in relation to occupational prestige, adjusting for other aspects of social class including household income and respondent's education. We analyzed data from 49,252 adults (age 25–64) from Cycle 2.1 of the Canadian Community Health Survey, a cross-sectional self-report survey conducted in 2003. Multiple linear regression was used to examine the relation between BMI and occupational prestige, adjusting for other sociodemographic variables. For women, higher ranking occupations showed lower average BMI relative to the lowest ranking occupations, but this effect was largely eliminated when adjusting for education. For men, occupation effects endured in adjusted models and we detected some evidence of a pattern whereby men in occupations characterized by management/supervisory responsibilities were heavier than those in the lowest ranking occupations (i.e., elemental sales and service). Results are interpreted in light of the symbolic value of body size in western culture, which differs for men and women. Men in positions of management/supervision may benefit from the physical dominance conveyed by a larger body size, and thus occupational prestige rankings may help us to understand the gender differences in the patterning of BMI by different indicators of social class.
The socioeconomic patterning of obesity and BMI in the general population remains poorly understood. Reviews of this topic indicate an inverse association (lower socioeconomic status, higher likelihood of obesity) among women in developed countries (1,2) that is most pronounced when education and occupation are used as indicators of socioeconomic status (2). For men in developed countries, the association is less consistent (1,2).
Exploration of these patterns is sparse in the health literature, which is unfortunate since understanding the socioeconomic patterning of weight—a uniquely visible physical attribute—may provide insight into socioeconomic inequalities in health in general. The social determinants of health literature highlights that inequalities in health are the result of inequalities in opportunities, resources, and constraints, which operate through an intersection of material and psychosocial mechanisms (3,4,5). Material mechanisms include income and the capacity to acquire resources such as adequate and acceptable food and appropriate housing in a safe neighborhood. Psychosocial mechanisms speak to issues of stress, control, and power that are a function of one's position in the social hierarchy. In the current global context, body weight is one manifestation of class-based inequality that clearly implicates these intersecting mechanisms, and thus provides an excellent illustration of the embodiment of inequality. Trends in food production and marketing associated with globalization have resulted in a surplus of processed, nutritionally poor food items that are more accessible to those of lower income, due to their lower cost (6,7,8,9). Likewise, labor market and broader social trends have contributed to an increasingly sedentary workforce (10,11), eliminating some occupation-based opportunity for physical exertion, and higher levels of leisure time physical activity among those of higher socioeconomic status (11,12) at least in part reflect better access and opportunity in these groups, such as the ability to afford fee-based activities such as gym memberships, and greater likelihood of living in proximity to pleasant and lower traffic areas for physical activity (11,13,14).
In terms of psychosocial dimensions with relevance for weight, a useful framework is the sociology of Bourdieu and his relational model of class (e.g., 15,16) as interpreted by others (e.g., 17,18,19). Of particular relevance is Bourdieu's concept of habitus, which refers to the embodiment of social structures in individuals (17). An individual's habitus is thought to be formed during childhood through sustained exposure to social structural axes (e.g., gender, class), and consequently predisposes an individual to behave in accordance with these social structures. Thus, the body (inclusive of appearance, style, and mannerisms) becomes a metaphor for one's status, and a visible bodily attribute such as weight becomes an ideal vehicle for the exploration of class inequalities. From this perspective, status has not just economic, but also cultural and social aspects (17), which can take on symbolic value when they are recognized as having prestige or legitimacy; for example, a particular accent, or a certain body size/shape, may have status or prestige that is not in keeping with its economic dimensions (17). Within an era characterized by pervasive commodification of the body, Shilling (19) highlights that people's identities become linked to the social values accorded to attributes such as size, shape, and appearance. Currently, and particularly for women in industrialized countries, there is a great deal of social value accorded to thinness, which is usually attributed to cultural values around beauty that are heavily promoted by western media (20,21,22). Based on this line of reason, it is plausible that a thinner body, along with related behaviors such as a healthy diet and regular physical activity, may be socially valued to a greater extent for those in higher social class (especially women in developed countries), in addition to being more materially viable in these groups. In support of this position, attributes such as dieting, pursuit of thinness, and body dissatisfaction have been shown to be more common among women of higher than of lower social class (23,24,25).
These processes, which have rarely been explored within epidemiological research on obesity, require the use of diverse indicators of class (26). Common indicators of class include income, education, and occupation/employment, to represent dimensions of theoretical importance such as structural location within the economy, material possessions and assets, and educational credentials or skills (26,27). In Canada, published research on the socioeconomic patterning of weight or obesity has been based primarily on income and education as indicators of class. Findings reveal a consistent inverse educational gradient in overweight and/or obesity for men and women (28,29,30,31,32,33,34,35,36,37), and a relation with income that is less straightforward, and that differs for men and women (32,37,38,39). Though substantive exploration of these associations is limited, the use of these indicators may privilege explanations of a material nature (26) and thus provide limited insight into aspects of social class suggested by a theory of class that includes habitus. On the other hand, examining associations with occupation—and particularly occupational prestige—may help us to tap into psychosocial and symbolic dimensions of class with relevance for understanding socioeconomic variation in BMI. While occupation-based indicators of social class are common in some countries (40,41), most notably in the United Kingdom (e.g., 42), this approach is less common in Canada. Although some studies have examined employment variables such as workforce participation (43), the role of ranked occupational hierarchical (i.e., prestige) as a dimension of class in relation to weight in Canadians is not well understood.
A small number of Canadian studies have examined obesity (or related variables) in relation to ranked occupational category (30,32,44), which collectively suggest a possible inverse relationship between obesity and this indicator of class; however there are significant limitations. The two population-based studies (30,32), which included ordered occupational groups such as unskilled, semiskilled, and professional, did not provide any background on these classifications used. Davis et al. (44) provided the basis for the occupation data used (i.e., the Blishen scale), but this study is limited by its use of a convenience sample of moderate size. None of these studies provided any substantive discussion about the possible mechanisms or pathways underlying the social patterning of weight, nor the implications unique to occupational class.
The objective of the present study was to examine the association between occupational prestige and BMI across a large, population-based sample of Canadian adults, adjusting for two more commonly used indicators of social class: income and education. It was hypothesized that an indicator of occupational prestige would better enable us to tap into the psychosocial and symbolic dimensions of class, and its role in explaining socioeconomic variation in BMI, above and beyond the role of indicators more commonly used in Canadian quantitative research: income and education.
Methods and Procedures
This work was based on Cycle 2.1 of the Canadian Community Health Survey (CCHS), a cross-sectional, nationally representative survey of the Canadian population 12 years and older conducted in 2003 (ref. 45). The CCHS covers ∼98% of the Canadian population; persons living on Indian Reserves or Crown lands, clientele of institutions, full-time members of the Canadian Forces and residents of certain remote regions are excluded (45). Households are selected using a multistage stratified cluster design that involves the area sampling frame used in the Canadian Labour Force Survey, and a list frame of telephone numbers to access most households. A small remaining proportion (2%) is accessed using random digit dialing. Respondents were interviewed using telephone and in-person administration modes, and an overall response rate of 80.7% was achieved (45). Data were accessed through the Research Data Centre program, a joint initiative of Statistics Canada, the Social Sciences and Humanities Research Council of Canada, and Canadian universities that exists to facilitate researcher access to Statistics Canada data.
Variables used in this study include self-reported height and weight (from which BMI, kg/m2, was computed), total household income adjusted for (divided by) the number of persons in the household (this variable was divided into quartiles), highest level of education achieved by the respondent (less than high school, high school graduate, some postsecondary, postsecondary degree or higher), occupational prestige ranking (described next), age, sex, marital status (married/common law; separated/divorced/widowed; and single), and usual number of hours worked per week.
To describe the occupation variable, some brief attention to the history of occupational measurement at the population level in Canada is helpful. In some of the earliest work on the topic, Pineo and Porter (46) developed prestige rankings for various occupations based on ratings of occupational titles made by respondents in a national survey. Because the prestige rankings were limited to the subset of occupations included in the survey, subsequent authors (e.g., 47,48,49) built on this work by incorporating average earnings and education associated with various occupations, so that a broader range of occupations could be classified. This work was considered to provide a composite measure of occupational status. In response to recent interest in reviving a prestige-based classification scheme, Canadian researchers Goyder and Frank (50) have updated the 1967 Pineo-Porter work to derive a prestige score for several occupational groupings, using a national survey approach that resembled the earliest work. Occupation data in CCHS 2.1 were classified using the 1991 Standard Occupational Classification (http:www.statcan.caenglishSubjectsStandardsoc1991soc91-index.htm), which were then recoded using concordance tables and formula into occupational prestige rankings (Table 1), via the National Occupational Classification for Statistics (http:www.statcan.caenglishSubjectsStandardsoc2001nocs01-index.htm).
Table 1. Frequency breakdown of rank-ordered occupational prestige groupings,a men and women age 25–64 (total n = 49,252)
All analyses were conducted using Stata/SE 8.2. Men and women were analyzed separately, and the sample was restricted to those of working age (25–64 years) who had complete data on age, education, income, occupation, marital status, and usual number of hours worked per week (n = 49,252). Analyses incorporated a probability weight to account for the complex sampling strategy used in the CCHS.
Following examination of descriptive statistics and bivariate associations, we used multiple linear regression to examine the association between BMI and occupational prestige rankings, adjusting sequentially for other sociodemographic variables (age, education, income, marital status, hours worked). Education, income, and occupation were entered as dummy variables, with the lowest category designated as the reference in all cases for ease of comparability. To improve distributional properties of the continuous outcome variable (BMI), values corresponding to a standardized score of >3.29 and less than −3.29 were eliminated as outliers (51) (this resulted in the elimination of 4.3% of the sample (n = 3,467), 87% of whom were women, and 99% of whom scored above 3.29 on the standardized BMI score). Our decision to focus on BMI as a continuous variable is based on our view of BMI as existing on a continuum, whereby low and moderate risk ‘merge imperceptibly’ into high risk (cf. 52), thus rendering categories and cutoffs somewhat arbitrary.
Of the total sample age 25–64, ∼30,700 individuals (37.7%) were eliminated due to missing data on variables used in this study (age, education, income, occupation, marital status, and usual number of hours worked per week). This could largely be attributed to missing data on three variables: income (23% missing data), occupation (20% missing data), and hours worked (20% missing data). To determine the pattern of missing data we compared mean (s.d.) BMI for those with and without data on these three variables. We found that those with missing data on income were lighter on average than those with complete data: M = 25.8 kg/m2 (s.d. = 4.9) vs. M = 26.0 kg/m2 (4.9) for those with missing and complete data, respectively (P < 0.01) (this pattern was also observed in men (P = 0.04) and women (P = 0.01) separately), and those with missing data on occupation and/or hours worked were heavier on average than those with complete data: M = 26.1 kg/m2 (5.5) vs. M = 25.9 kg/m2 (4.8) for missing vs. complete data on occupation and M = 26.1 kg/m2 (5.5) vs. M = 25.9 kg/m2 (4.8) for missing vs. complete data on hours worked (these patterns were also observed in women separately (P < 0.01 in both cases). The pattern for men was in the same direction but was nonsignificant). Although this introduces an element of bias into our subsample, it enables the analyses to be directly comparable to one another, which is important when examining the results of multivariate analyses with sequential adjustment from one model to the next.
Descriptive statistics for age, BMI, hours worked, marital status, income, and education are presented in Table 2. Men on average were older, had a higher BMI, and worked more hours per week, than women (P < 0.001). Frequency breakdown for the ranked occupational prestige categories are presented in Table 1. The most common job category for men was trades and skilled transport and equipment operator (16%) followed by middle and other management occupations (10.7%). For women, clerical occupations were most common (14.4%), followed by intermediate sales and service (11.6%) and skilled administration and business occupations (10.9%). The association of age with BMI was predominantly linear in nature, as illustrated by age group mean values in Table 2.
Table 2. Descriptive statistics for study variables, men and women age 25–64 (total n = 49,252)
Based on analysis of variance, an inverse association between BMI and education was observed for both men (F(3,26330)=122.2, P < 0.001) and women (F(3,22914)=155.5, P < 0.005). For men, those with complete postsecondary training had a lower BMI than those with less than high school, and for women, those with complete high school, some postsecondary training, or complete postsecondary training had a lower BMI than those with less than high school (Figure 1a). For income on the other hand, an inverse association was observed for women (F(3,22914)=25.6, P < 0.001) whereby those in the top two income quartiles had lower BMI than those in the lowest income quartile, while a positive association was observed for men (F(3,26330)=16.7, P < 0.001) whereby those in the top three income quartiles had higher BMI than those in the lowest income quartile (Figure 1b).
Results of multiple linear regression
For women (Table 3), associations with occupational prestige, adjusting for age only (model 1), indicate that relative to the reference category (elemental sales and service, which includes jobs such as cashiers and janitors), many occupations had a lower average BMI. This was especially true for occupations near the top of the prestige rankings: seven of the top ten occupations showed a lower average BMI relative to elemental sales and service in model 1: professional occupations in health (e.g., physicians, pharmacists); technical and skilled occupations in health (e.g., medical laboratory technicians; midwives); professional occupations in social science, education, government service, and religion (e.g., professors/teachers, social workers); professional occupations in natural and applied sciences (e.g., engineers, architects); technical occupations related to natural and applied sciences (e.g., conservation and fishery officers, air pilots); professional occupations in business and finance (e.g., accountants, investment dealers); and professional occupations in art and culture (e.g., curators, writers). Women in middle and other management occupations (e.g., financial managers; managers in health care) also showed a lower BMI, on average, than the reference category, in model 1. While this pattern of findings was largely retained when further adjusting for income (model 3), nearly all effects were eliminated when adjusting for education (model 2), suggesting that effects of occupational prestige are largely attributable to education level. No effects were apparent in the fully adjusted model (model 4).
Table 3. Results of multiple linear regression analysis examining BMI in relation to occupational prestige for men (n = 26,334) and women (n = 22,918), systematically adjusting for age, education, income, marital status, and usual hours worked per week
For men (Table 3), associations adjusting for age only (model 1) resembled a nonlinear pattern whereby those in certain higher prestige jobs (e.g., professional occupations in health) were lighter on average, and those in certain mid-range prestige jobs (e.g., skilled administrative and business occupations; intermediate occupations in transport, equipment operators, installation and maintenance) were heavier on average, than those in the reference category of elemental sales and service. Adjusting for education or income, several effects remained and/or emerged, and for parsimony model 2 shows effects adjusted for both income and education. In these analyses, the following occupational groups showed a higher average BMI than the reference category: senior management occupations (e.g., senior government managers and officials, senior managers in health), skilled administrative and business occupations (e.g., administrative officers, conference and event planners), middle and other management occupations (e.g., financial managers, managers in health care), and intermediate occupations in transport, equipment operators, installation and maintenance (e.g., truck drivers, pest controllers). Workers in technical and skilled occupations in art, culture, recreation and sport (e.g., photographers, athletes) were lighter on average than the reference category. In the fully adjusted model (model 3), most effects were eliminated, with the exception of lower average BMI among workers in professional occupations in health, and technical and skilled occupations in art, culture, recreation and sport, and higher average BMI among workers in intermediate occupations in transport, equipment operators, installation and maintenance. Moving from model 2 to model 3, additional regression analyses adjusting for one covariate at a time revealed no clear pattern of explanation by either marital status or hours worked.
The aim of this study was to further probe the socioeconomic patterning of weight among Canadian adults, by exploring the role of occupational prestige. Whereas other studies in this vein have focused on income and education as indicators of class, or on other occupational attributes such as workforce participation, job strain, and working conditions on weight (43,53,54), our approach was to use occupational prestige as a means of tapping into psychosocial and symbolic dimensions of socioeconomic differences in BMI, drawing on Bourdieu's theoretical model of class. Overall, we detected negligible independent effects of occupational prestige on BMI for women. In models adjusting just for age, and for age and income, women in skilled professional occupations tended to be thinner on average than those in lower ranking occupations, and it is certainly plausible that the workplace culture in higher ranking occupations is one in which thinness in particular and physical appearance in general are valued. However, the elimination of most of these effects in other adjusted models suggests that education is also important in this regard. Education may be a marker for status-related attributes such as interest in and ability to comprehend health promotion messages, as well as investment in one's body and health. High levels of education may imply a degree of achievement aspiration that manifests not just in the education domain, but more broadly as a desire to achieve a body type with the greatest symbolic value—which for women, is thin (23). Our findings may also speak to a lesser prominence of occupation for defining class or status for women vs. men (e.g., 26,55); for example, research has highlighted that while occupation and associated prestige and income are key drivers of status and social mobility in men, these have traditionally been regarded as less relevant for women, for whom mobility through marriage was commonplace (55). Relatedly, women have historically been more likely than men to be underplaced occupationally according to their education, reflecting in part a more limited range of jobs available, and a general phenomenon of lower occupational and economic payoff for women than for men of comparable academic achievement (55). Though these trends may be changing, they provide some support for our position that other indicators of class (such as education) may better tap into those aspects of social class that have relevance for women's weight and health.
For men on the other hand, occupation effects were more robust to adjustment for income and education. In the fully adjusted models, a lighter body size on average was observed among men in professional health-related occupations (category 1), and in technical and skilled occupations in art, culture, recreation and sport (category 13). These findings probably reflect some combination of the activity level inherent in these occupations (for example, health professionals in category 1 may be on their feet a great deal; category 13 includes athletes), and the focus on health and/or recreation in these categories, which might indicate a selection effect of thinner workers (however, other health-related occupational categories did not show this effect, such as technical and skilled occupations in health, and assisting occupations in support of health services). For professional occupations in health, the prestige of these occupations may also contribute, via a social norm in which health, fitness, and a slim body are valued and encouraged. In the models adjusting only for the two other class variables (income and education), a finding of interest was the heavier body size among those in senior and middle management occupations. Looking collectively at research on the socioeconomic patterning of weight in Canadian men, we see an inverse association with education (e.g., 33; the present research), but an effect tending toward positive for income (39; the present research) and occupation (the present research). This constellation of findings merits some consideration.
The inverse BMI-education association in men may to some extent reflect similar pathways as in women, including greater comprehension of and ability to comply with health promotion messages around food and exercise, as well as a desire to optimize health within a culture where obesity continues to be highly stigmatized (56,57,58). However, we are left with the seemingly contradictory finding of an association of opposite direction with income and occupation. Drawing again on the work and interpretations of Bourdieu (15,16,17,18), we highlight that the dimensions of the valued body differ between the genders. In the face of stigma associated with excess weight that affects both men and women, men are also exposed to a conflicting social pressure: valuation of a larger body size as conveying strength, power, and dominance. This is illustrated, for example, in research on gender differences in body image among children: McVey et al. (59) have shown that while girls wish to be thinner, boys—despite societal discrimination against obesity—often wish to be larger (and in particular, more muscular). Further, weight-based discrimination appears to be less prominent for men than women, based on research documenting economic consequences of obesity (e.g., wage penalty) for women but not men (56,60). Thus, social pressures, which help to define the symbolic value of body size, are consistent for women (valuation of thinness by all accounts) but are somewhat conflicting for men. For men who are in higher ranking occupations—especially those that involve management and supervisory responsibilities, a larger body size (whether overweight, muscular, or some combination) could be valuable toward asserting one's dominance or authority. Further, with men being the traditional earners in families (another form of dominance or authority), it is plausible that income and pursuit of a larger body size remain linked in a similar manner. Another explanation for gender differences in the income-weight relationship concerns our use of a household-level indicator of income. Prospective longitudinal studies (61,62) have demonstrated a marriage market penalty for heavier women, whereby obese women (but not men) are less likely to marry. Obese women who do marry are likely to have spouses with lower earnings than non obese women, adjusting for various baseline health and social variables. Thus one might speculate that—among married heterosexual couples in the higher income range who rely mainly on the male income—a heavier body size for men is more acceptable, whereas for women a heavier body size may put the marriage at risk. One consideration that remains for the occupation findings in men is that it is the sedentary nature of these jobs that drives the larger average body size (due to lack of occupation-based physical activity). However, this would seem unlikely for two reasons: first, other highly ranked occupational categories that are equally sedentary (e.g., professional occupations in government services, natural and applied sciences) did not show this effect. Second, even if men in these higher ranking occupations are not accumulating much occupational physical activity, existing literature would indicate that they are still more likely than their lower status counterparts to engage in physical activity in their leisure time (11,12).
The use of self-reported height and weight is a key limitation to this study, and it would be valuable to replicate results using data based on measured values. Unfortunately, the only recent Canadian population-based dataset that includes measured height and weight (CCHS Cycle 2.2) does not include detailed occupational information and was therefore not suited to answering these research questions. Limitations of cross-sectional research, including the inability to attribute temporality, also apply here. Behavioral mediators such as diet, which have been explored elsewhere (ref. 18,63 and J. Godley and L. McLaren, unpublished data), were not a focus here, and would be one important pursuit for further research in Canada and elsewhere. In terms of next steps, it would be valuable to pursue alternative conceptualizations of occupational class, such as those focusing on relations of power and dominance, to explore whether implications for weight and health are different (64). In general, perhaps the key challenge for future research on this theme is to refine the conceptualization(s) of social class, including the intersection of various markers of social class, and to work out the implications for measurement and analysis in population-based survey research.
Although the research and analysis are based on data from Statistics Canada, the opinions expressed do not represent the views of Statistics Canada. L.M. gratefully acknowledges the support of a Population Health Investigator Award and Establishment Grant from the Alberta Heritage Foundation for Medical Research (AHFMR). We wish to acknowledge Richard A. Wanner, Professor in the Department of Sociology, University of Calgary, for his assistance with this project.