Socioeconomic status and body mass index in Canada: exploring measures and mechanisms.
Godley, Jenny ; McLaren, Lindsay
BACKGROUND
Socioeconomic Status (SES) and Health
THERE IS SUBSTANTIAL EVIDENCE OF SOCIOECONOMIC inequalities in
health in both social science and medical literatures. Health
researchers in Europe and North America have repeatedly shown that there
is a relationship between SES and mortality, morbidity, health
behaviors, and access to and utilization of health-care resources, in
favor of those with higher SES (Adler and Ostrove 1999; Commission on
the Social Determinants of Health 2008; Feinstein 1993; Frohlich, Ross,
and Richmond 2006; House 2001; Link and Phelan 1995; Mackenbach et al.
2008; Ross and Wu 1995).
In an attempt to reduce such health disparities, several
industrialized countries, including Canada, developed nationalized
health-care systems in the middle of the last century. Evaluations of
these systems have repeatedly demonstrated that they have failed to
eliminate socioeconomic differences in health outcomes, behaviors, and
utilization. In the United Kingdom, for example, both the 1982 Black
report and the 1992 Whitehead report showed that SES remained closely
linked with health status, decades after the implementation of the
National Health Service (Townsend, Davidson, and Whitehead 1992).
Subsequent research has shown that these inequalities persist, and may
in fact have worsened (Chandola and Jenkinson 2000). In Canada, much
health research highlights the fact that despite universal access to
health care, socioeconomic disparities in health remain evident both
within and across provinces and territories (Frohlich, Ross, and
Richmond 2006; Humphries and van Doorslaer 2000; Poudrier 2007; Prus
2007; Raphael et al. 2006). There is a clear need for researchers to
consider mechanisms other than access to health care when attempting to
understand social inequalities in health in Canada.
Understanding and modeling the mechanisms through which SES affects
health remains a challenge (Raphael et al. 2006; Scambler and Higgs
1999; Tugwell and Kristjansson 2004). In the United Kingdom, the Black
report originally posited that socioeconomic disparities in health could
be understood using two main types of explanations: cultural/behavioral
explanations (e.g., social class differences in health behaviors) and
materialist/structuralist explanations (e.g., social class differences
in access to health promoting resources) (Townsend et al. 1992). More
recent studies in the United States have focused on a third type of
explanation--social class differences in psychosocial experiences and
resources (e.g., social class differences in the experience of stress or
the sense of personal control) (Aneshensel 1992; House 2001; Lantz et
al. 2005).
The list of mechanisms through which SES may affect health can thus
be summarized as material, cultural, and psychosocial (Prus 2007). (1)
All of these mechanisms have been shown to mediate the relationship
between different measures of SES and health (Borrell et al. 2004;
Frohlich, Ross, and Richmond 2006; Rahkonen et al. 2006). Different
researchers focus on different mediators, depending on both the way they
operationalize socioeconomic position, and the dependent variable or
health outcome they are examining.
In this paper, we explore material and cultural explanations for
the relationship between SES and body mass index (BMI). The measurement
and operationalization of SES is often debated in the health literature
(Krieger, Williams, and Moss 1997; Veenstra 2007). We examine education
measured at the individual level and income measured at the household
level. We argue that the effects of education usually accrue to the
individual who experiences the educational setting, thus an
individual-level education variable is appropriate. However, household
income has been shown to be a better proxy measure of social class than
individual income, especially for women (Krieger, Chen, and Selby 1999;
Macintyre and Hunt 1997).
We are particularly interested in focusing on education and income
separately by gender as there have been contradictory findings regarding
the impact of these indicators of SES on BMI (Ball, Mishra, and Crawford
2002; Kuhle and Veugelers 2008; McLaren 2007; McLaren and Godley 2009).
These differences by gender and by indicator of SES (outlined below) may
implicate different mechanisms in the processes which produce
inequalities in health. While the effect of education on health may be
more likely to reflect cultural factors, the effect of income on health
may be more likely to operate through material factors. It is admittedly
difficult to tease apart (or even classify) the types of mechanisms at
work, but we speculate on what the different results for education and
income may tell us about the mechanisms in our discussion.
In this paper, we focus on lifestyle mediators of the relationship
between SES and BMI. This focus follows naturally from our dependent
variable, as there are well-established relationships between certain
health and lifestyle behaviors (such as diet, exercise, smoking, and
alcohol consumption) and body weight (Kuhle and Veugelers 2008; Ricciuto
and Tarasuk 2007). The behavioral and lifestyle variables we examine,
and their potentially differential association with income and
education, may provide insight into both material and cultural
mechanisms through which SES affects health. (2) Psychosocial mechanisms
are harder to capture in the Canadian Community Health Survey (CCHS)
data. In our discussion, we speculate that some of the socioeconomic
variance in BMI that remains unexplained by our models may represent
unmeasured psychosocial factors.
BMI
The emergence of the obesity "epidemic" in the developed
world over the past 30 years is well-documented (Katzmarzyk 2002;
Shields and Tjepkema 2006; Tremblay, Katzmarzyk, and Williams 2002;
World Health Organization 1997). Recent research in Canada based on
measured weight and height shows that 23 percent of the general
population is clinically "obese" and an additional 36 percent
is "overweight" (Tjepkema 2006). Obesity has been linked to
many negative physical health outcomes, including heart disease and
diabetes, and mental health outcomes, such as depression and mortality
(McLaren et al. 2008; Rashad 2003; Ross et al. 2007; Tjepkema 2006). The
increased obesity prevalence has medical, labor market, and social
consequences, and is thus cause for great concern among health and
social science researchers (Klarenbach et al. 2006; Peralta 2003).
Importantly, body weight is not simply a "health
outcome." Body weight is also a personal characteristic that has
tremendous sociological significance. Personal appearance, including
body weight, affects many aspects of our social lives including how we
present ourselves to others, how we treat others, and how we are treated
(Carr and Friedman 2005; Goffman 1959). Using Bourdieu's (1984)
notion of "habitus," we suggest that an individual's
social status is both written on and reflected through the body which
she/he inhabits and presents to the world. The body as
"habitus" forms an integral part of one's social status,
as do attempts to control or change the body, including behaviors that
affect weight, such as diet and exercise (Power 1999; Warin et al.
2008).
One's "habitus" is highly gendered. There is a long
history of feminist scholarship examining the social value placed on
women's appearance and the gendered nature of the social and
psychological correlates of appearance, appearance-related behaviors,
and appearance-related illnesses (including eating disorders and body
image disturbance) (Bookwala and Boyar 2008; Bordo 1993; Clarke 2001;
Frost 2001; Orbach 1979, 1986; Pipher 1995; Price and Shildrick 1999).
Recently, this literature has expanded to focus on the increasing
significance of appearance for men (Grogan 1999; Luciano 2001; Pope,
Phillips, and Olivardia 2000; Swami et al. 2008). There are gender
differences in the social statuses assigned to various aspects of
appearance; for example, significant social value is placed on thinness
for women in the developed world, while a larger and more muscular
physique is considered desirable for men (Grogan 1999; McVey, Tweed, and
Blackmore 2005). The relationships between different aspects of SES and
weight likely reflect processes related to this differential social and
symbolic value of body size in our society (McLaren and Godley 2009;
Peralta 2003). Thus, examining the mechanisms through which SES affects
body weight is not just a study of the reproduction of health
inequalities; it is an investigation into the reproduction of social
inequality itself.
SES and BMI
International research has documented a link between obesity and
SES. Sobal and Stunkard, in their literature review in 1989, and McLaren
in her updated literature review in 2007 demonstrate that while
education has been consistently shown to be negatively related to body
weight for both men and women in the developed world, results for income
are less consistent. There have been several studies of Canadian adults,
using various data sets, illustrating socioeconomic differences in BMI
(Cairney and Wade 1998; Ostbye et al. 1995; Shields and Tjepkema 2006;
Willms, Tremblay, and Katzmarzyk 2003). In Canada, researchers using
education as their primary measure of SES have found that for both men
and women, higher levels of education are correlated with lower BMI.
However, research based on income has shown mixed results. Using data
from the 1994 National Population Health Survey, Cairney and Wade (1998)
found no significant relationship between income and obesity. Recent
data from the Cycle 2.2 (CCHS 2.2) shows a nonlinear relationship
between income and BMI for women, and a positive relationship for men
(Shields and Tjepkema 2006).
Additional sociodemographic variables that have been shown to be
correlated with BMI in the industrialized world include age, marital
status, and race/ethnic origin. BMI increases with age up until the
older years (Prus 2007). Married men and women tend to be heavier than
unmarried men and women (Borders, Rohrer, and Cardarelli 2006; Rashad
2003; Ross et al. 2007). Research in both Canada and the United States
demonstrates significant racial and ethnic differences in the prevalence
rates for obesity, even controlling for age, education, income, birth
place, and physical activity levels (Borders et al. 2006;
Sanchez-Vaznaugh et al. 2009; Tremblay et al. 2005); for example,
Aboriginals consistently shows an obesity prevalence that is higher than
average (Poudrier 2007), whereas East Asians tend to have lower than
average prevalence. While these variables are not the focus of the
current study, we control for all of these sociodemographic variables in
our multiple regression models.
Diet and exercise are the primary behavioral determinants of BMI
(3) (Ball, Mishra, and Crawford 2003; Brien et al. 2007; Janssen et al.
2006), though smoking and alcohol intake are also relevant (Birch et al.
2005; Cairney and Wade 1998). Several previous studies have examined
socioeconomic differences in these proximate determinants of obesity
(Hall et al. 2003; Kuhle and Veugelers 2008; Matheson, Moineddin, and
Glazier 2008; Peralta 2003; Power 2005; Trovato and Lalu 2007; Ward,
Tarasuk, and Mendelson 2007). In Canada, one study found that women of
higher SES reported more physical activity and higher fruit and
vegetable intake, which helped explain their lower obesity risk (Kuhle
and Veugelers 2008). Other Canadian studies have reported that higher
income men reported lower likelihood of smoking (Ward et al. 2007),
greater likelihood of smoking cessation (Kuhle and Veugelers 2008), and
less physical activity (Kuhle and Veugelers 2008), which helped to
explain their higher obesity risk. We build on this work by casting a
sociological lens on the interrelationships between income, education,
and a broad range of potential sociodemographic and lifestyle variables,
moving toward understanding material and cultural explanations for the
gendered socioeconomic patterning of body weight.
Using nationally representative data on working age adults from the
CCHS 2.1, we pose three research questions: Does the relationship
between BMI and SES vary by gender? Does the relationship between SES
and BMI remain once we control for sociodemographic variables? Can the
relationship between BMI and SES, net of sociodemographic variables, be
explained by behavioral and lifestyle variables? Our results help
elucidate the mechanisms through which socioeconomic inequalities in
health are established and reinforced in the Canadian context.
METHODS
Data
We analyze data from the CCHS 2.1. This survey, conducted in 2003,
is a cross-sectional, nationally representative survey of the Canadian
population. The target population is people aged 12 and older living in
private dwellings in the 10 provinces and three territories. The CCHS
sampling frame covers approximately 98 percent of the Canadian
population; those excluded are: people living on Indian Reserves or
Crown lands; people who are institutionalized; flail-time members of the
Canadian Forces; and people living in certain remote regions (Statistics
Canada 2005).
Households are selected using a multistage stratified cluster
design, based on the area sampling frame devised for the Canadian Labour
Force Survey. A list of telephone numbers was used to reach most
respondents; the remaining 2 percent were accessed using random digit
dialing. Both in-person and telephone interviews were conducted, using
computer-assisted interviewing software. The overall response rate for
the CCHS 2.1 was 80.7 percent (Statistics Canada 2005).
Data were accessed at the Prairie Regional Research Data Centre at
the University of Calgary. All analyses were conducted using Stata/SE
9.0. All analyses incorporated sample weights, as provided by Statistics
Canada. The standard errors in the regression models were estimated
using a bootstrap procedure, using the 500 bootstrap weights provided by
Statistics Canada, to account for the complex sampling design used in
the CCHS 2.1 (Pierard, Buckley, and Chowhan 2004).
We restrict our analyses to respondents aged 25 to 64 for whom we
have self-reported weight and height data. We focus on the working age
population as this is the group for whom individual-level education and
household-level education are most comparable. To improve distributional
properties of the continuous outcome variable (BMI, described below), we
eliminate respondents who have a standardized BMI score of > 3.29 and
< -3.29 (4) (Tabachnick and Fidell 1996). Our full sample size is
77,499 (37,578 males and 39,921 females). Men and women were analyzed
separately.
Measures
BMI (kg/[m.sup.2]) was computed from self-reported weight and
height. In this paper, we focus on the continuous measure of BMI rather
than using a categorical variable with categories that delineate
"overweight" and "obese." We adopt Rose's
(1992) view that BMI, like many other physiological characteristics, is
better represented as a continuous variable. Using somewhat arbitrary
cutoffs to create categories out of this continuum may misrepresent the
extent of risk indicated by increases along the spectrum of BMI.
Additionally, we are constrained by the fact that the height and weight
data in the CCHS 2.1 are self-reported. While some research has shown
that self-reports may underestimate BMI compared with measured height
and weight data (John et al. 2006), others have argued that
self-reported BMI is valid for use in epidemiological studies as long as
it is used as a continuous, and not a categorical, variable (Spencer et
al. 2002).
Highest level of education achieved by the respondent was recorded
in 10 categories, and recoded into four categories: less than high
school; high school graduate; some postsecondary; university degree or
higher. These categories were created to reflect substantively
meaningful educational attainment within the Canadian school system:
primary, secondary, and tertiary schooling (Borders et al. 2006;
Tremblay et al. 2005).
Following Humphries and van Doorslaer (2000), total household
income was adjusted for household size by dividing by the square root of
the household size (thus accounting for economies of scale in larger
households) (Mackenbach et al. 2008). (5) In preliminary analyses, we
examined the bivariate association between household income measured as
a continuous variable and BMI stratified by sex, and found that it was
not linear, in particular for men. To capture this nonlinearity we opted
to analyze income as a categorical variable. Household income was
divided into quartiles; mean values of total household income for the
four quartiles were $23,912, $50,541, $75,200, and $132,701,
respectively. Mean household size was 3.1 for the first three income
quartiles and 2.8 for the highest quartile. The distribution of the two
SES variables, separately by gender, is shown in Table 1.
We note that 1,362 respondents are missing data on education and
16,568 respondents are missing data on income. Average BMI did not
differ for men and women with and without data on the SES variables.
The sociodemographic control variables were measured as follows.
Age, reported in years, was treated as a continuous variable.
Cultural/racial background was based on respondents' report of
their "cultural/racial origin." They could choose from the
following options: white, black, Korean, Filipino, Japanese, Chinese,
Aboriginal, South Asian, Southeast Asian, Arab, West Asian, Latin
American, and Other. Those who chose more than one option were
categorized as "Multiple Origin." Based on previous research
on racial/ethnic differences in body weight in the United States and
Canada (Sanchez-Vaznaugh et al. 2009; Tremblay et al. 2005), we derived
six categories for cultural/racial background from this data: white;
black; East Asian; South Asian; Aboriginal; and Other (which included
Latin American, Other, and Multiple Origin). Marital status was
categorized as follows: married or common law; separated or divorced;
single; widowed. The distribution of the sociodemographic control
variables, by gender, is shown in Table 1.
The distribution of the lifestyle variables is also shown in Table
1. Physical activity was measured as the number of times the respondent
reported engaging in vigorous physical activity for 15 minutes or more
over the past month. Dietary intake was measured as the total servings
of fruit and vegetables (excluding potatoes) respondents reported eating
per day. Smoking frequency was categorized as daily, occasionally, or
not at all. Alcohol consumption was categorized into four categories:
drink less than once a month; drink up to three times a month; drink one
to three times a week; drink four or more times a week. Finally, hours
worked refers to the usual number of hours worked in the last week at
one's main job or business. Following Artazcoz et al. (2009), we
created four groups: part time (29 hours or less a week); full-time
(30-40 hours a week); slightly over full-time (41-50 hours a week); and
greatly over full-time (51 or more hours a week). We also included a
category for those who did not work for pay.
Analyses
To answer our first research question, Does the relationship
between BMI and SES vary by gender?, we ran linear regression models,
regressing BMI on education and income separately by gender. To answer
our second research question, Does the relationship between SES and BMI
remain once we control for sociodemographic variables ?, we reran these
linear regression models, controlling for age, marital status, and
cultural/racial background. To answer our third research question, Can
the relationship between BMI and SES, net of sociodemographic variables,
be explained by behavioral and lifestyle variables ?, we incorporated
our lifestyle variables into the full multiple linear regression models
of BMI.
RESULTS
Tables 2 and 3 present the results of the regression models. Each
model was run separately, including only those respondents for whom we
had complete data on all the variables in the final model ("Model
C"). We first discuss the results for income and BMI (Table 2), and
then the results for education and BMI (Table 3).
In Table 2, "Model A," we examine the bivariate
relationship between income and BMI. We find that income is inversely
related to BMI for women. There is a positive relationship between
income and BMI for men aged 25 to 64, but this relationship is not
linear. Men in the highest income category have significantly higher
BMIs than men in the lowest income category, but there are no BMI
differences among men in the middle income quartiles.
The column for "Model B" in Table 2 shows the impact of
income on BMI adjusting for sociodemographic controls, and the column
for "Model C" includes the lifestyle variables. Summarizing
these findings, the relationship between income and BMI remains inverse
and significant (and in fact gets stronger) among women adjusting for
sociodemographic controls.
For men, the relationship between income and BMI remains positive
for the comparison between the highest and lowest income men, but this
relationship is reduced in magnitude with the inclusion of the
sociodemographic control variables.
With the inclusion of the behavioral variables ("Model
C"), the effects of income on BMI for women are reduced to below
the initial bivariate values. Thus, we see that the negative effect of
income on BMI for women is partially, but not fully, explained by the
lifestyle and behavioral variables. It appears that higher income women
engage in behaviors that contribute to lower BMI, including exercising
more and eating more fruits and vegetables than lower income women.
For men, the positive effect of income on BMI is reduced to
non-significance with the inclusion of the behavioral variables. Thus
for men, we see that the effect of income is fully mediated by the
combination of sociodemographic and lifestyle variables. These results
indicate that it is the lifestyle that goes along with earning a higher
income (including exercising less and working longer hours) which is
contributing to higher income men's higher BMI.
Examining Table 3, "Model A," we find that education is
inversely related to BMI for men and women aged 25 to 64. The column for
"Model B" shows that the relationship between education and
BMI remains negative for both men and women, across all four education
categories, controlling for the sociodemographic variables, although
there is some reduction in the magnitude of effect.
The results shown for "Model C" (which includes the
lifestyle variables) demonstrate that for men, the effect of education
is actually stronger in the full models. Once we control for smoking,
alcohol consumption, hours worked per year, and fruit and vegetable
consumption, men with university degrees are predicted to have even
lower BMIs than those with lower levels of education. Far from
explaining the effect of education on BMI for men, the inclusion of the
lifestyle variables has actually strengthened the effect. Lifestyle
variables which suppress the effect of education for men include smoking
and hours worked (higher education men smoke less and work longer hours
than lower education men, both of which are positively correlated with
BMI).
Comparing the education coefficients in the female models, we see
that the effect of education on BMI for women is slightly reduced with
the inclusion of the lifestyle variables. Thus, part of the reason that
higher education women have lower BMIs is that they are more likely to
exercise and eat more fruit and vegetables than lower education women.
DISCUSSION
Our results highlight the complexity of the relationship between
SES, behavioral factors, and BMI. In response to our first two research
questions (i.e., Does the relationship between BMI and SES vary by
gender? Does the relationship remain once we control for
sociodemographic variables?), we find evidence that the relationship
between BMI and SES, controlling for sociodemographic variables, does
vary by the measure of SES used and by gender.
Overall, education is more strongly and consistently related to BMI
for both men and women than income. Models including education and
sociodemographic controls explain slightly more of the variation in BMI
than models including income and sociodemographic controls. Education
has an inverse relationship for women, both in the bivariate results and
in the models controlling for sociodemographic variables. Education also
has an inverse relationship with BMI for men, but explains slightly less
of the variance in BMI than it does for women.
Income also has a consistently inverse relationship with BMI for
women, controlling for sociodemographic variables. For men, though,
income has a nonlinear positive relationship with BMI, with men in the
highest income quartile having a higher BMI than men in the lowest
income quartile, controlling for age, marital status, and
cultural/racial background.
These findings are consistent with what others have found when
examining the relationship between BMI, education, income, and gender in
the developed world (Kuhle and Veugelers 2008; McLaren 2007). These
findings suggest that cultural factors (as represented by educational
attainment) may be more important than material factors (as represented
by income) in understanding social class disparities in BMI and in
health outcomes more generally (Prus 2007). Additionally, these findings
add support to the theory that body weight (as part of an
individual's "habitus") may be a more important status
indicator for women than for men in Canadian society (Clarke 2001;
Peralta 2003).
In response to our third research question (i.e., Can the
relationship between BMI and SES, net of demographic variables, be
explained by behavioral and lifestyle variables?), we find mixed
results. For women, the inverse effects of both education and income on
BMI are partially explained by fruit and vegetable consumption,
exercise, smoking, alcohol consumption, and hours worked. For men, the
positive effect of income on BMI is fully explained by these variables.
However, for men the negative effect of education on BMI is actually
strengthened by the inclusion of the behavioral variables.
These findings clarify two previous findings in the literature
concerning the complex relationship between SES and BMI. Others have
noted the seemingly contradictory finding that income is positively
related to BMI for men (Shields and Tjepkema 2006; Zhang and Wang 2004).
Once we control for demographic variables, we find that the positive
relationship between income and BMI for men is completely mediated by
the lifestyle variables.
Others have also noted that the relationship between education and
BMI is stronger for women than for men (Shields and Tjepkema 2006;
Willms et al. 2003). Our bivariate models replicate this finding.
However, in our more complex models, we find that lifestyle variables
mediate the relationship between education and BMI for women and
actually suppress the relationship for men. Once we account for fruit
and vegetable consumption, exercise, smoking, alcohol consumption, and
hours worked, the effects of education on BMI are very similar for women
and men.
Our analyses highlight the importance of considering an array of
behavioral and lifestyle factors as part of the gendered nature of the
body as "habitus." Higher status women (measured both by
education and income) engage in a variety of behaviors that contribute
to lower body weight, not all of which are healthy. Higher status men,
on the other hand, engage in some behaviors that contribute to higher
body weight, such as exercising less and working longer hours, than
lower status men. Thus the "habitus" of higher status men may
also be unhealthy, in a way that contributes to higher, not lower, body
weight. "Habitus," therefore, may have implications for other
health outcomes.
Limitations of our analyses include the fact that we are reliant on
self-reported weight and height in the CCHS 2.1. Additionally, because
we are using cross-sectional data, we can not draw causal conclusions.
Future research could use longitudinal data to examine the mechanisms
through which social status affects health behaviors and BMI across the
life course.
Nevertheless, our results help elucidate the mechanisms through
which socioeconomic inequalities in health are established and
reinforced in the Canadian context. We find that the measure of SES
matters. Once we control for sociodemographic and lifestyle variables,
income shows a reduced effect on BMI for women and no effect for men.
Education, however, shows a consistent, negative effect on BMI for both
women and men, controlling for behavioral variables. Thus we conclude
that the effect of education on BMI must operate through more than just
cultural and material factors, which we have operationalized through
health behaviors and lifestyle. Future research should examine other
mechanisms, such as psychosocial factors, to explain the education-BMI
link. Examples include psychosocial stress, experiences of power and
hierarchy, and perhaps social comparison, among others.
We conclude by asserting that inequalities in BMI reflect both
material and cultural, as well as psychosocial mechanisms, and that it
is important to operationalize and explore these separate yet
overlapping pathways. Because education remains a stronger predictor of
BMI than income for both genders, particular effort toward understanding
the cultural mechanisms (that are not captured in the sociodemographic
and lifestyle variables examined here) is warranted. Inequalities in BMI
also highlight the importance of the body as both an integral element of
and a reflection of one's "habitus" (Bourdieu 1984). We
find that inequalities in BMI are stronger for women than men,
suggesting that body weight remains a gendered indicator of social
status.
References
Adler, N.E. and J.M. Ostrove. 1999. "Socioeconomic Status and
Health: What We Know and What We Don't." Annals of the New
York Academy of Sciences 896:3-15.
Aneshensel, C.S. 1992. "Social Stress: Theory and
Research." Annual Review of Sociology 18: 15-38.
Artazcoz, L., I. Cortes, V. Escriba-aguir, L. Cascant and R.
Villegas. 2009. "Understanding the Relationship of Long Working
Hours with Health Status and Health-Related Behaviours." Journal of
Epidemiology and Community Health 63:521-27.
Ball, K., G.D. Mishra and D. Crawford. 2002. "Which Aspects of
Socioeconomic Status Are Related to Obesity among Men and Women?"
International Journal of Obesity and Related Metabolic Disorders
26:559-65.
Ball, K., G.D. Mishra and D. Crawford. 2003. "Social Factors
and Obesity: An Investigation of the Role of Health Behaviors."
International Journal of Obesity and Related Metabolic Disorders
27:394-403.
Birch, S., M. Jerrett, K. Wilson, M. Law, S. Elliott and J. Eyles.
2005. "Heterogeneities in the Production of Health: Smoking, Health
Status and Place." Health Policy 72:301-10.
Bookwala, J. and J. Boyar. 2008. "Gender, Excessive Body
Weight, and Psychological Well-Being in Adulthood." Psychology of
Women Quarterly 32:188-95.
Borders, T.F., J.E. Rohrer and K.M. Cardarelli. 2006.
"Gender-Specific Disparities in Obesity." Journal of Community
Health 31:57-68.
Bordo, S. 1993. Unbearable Weight: Feminism, Western Culture and
the Body. Berkeley, CA: University of California Press.
Borrell, C., C. Muntaner, J. Benach and L. Artazcoz. 2004.
"Social Class and Self-Reported Health Status among Men and Women:
What Is the Role of Work Organization, Household Material Standards and
Household Labour?" Social Science and Medicine 58:1869-88.
Bourdieu, P. 1984. Distinction: A Social Critique of the Judgement
of Taste. London: Routledge.
Brien, S.E., P.T. Katzmarzyk, C.L. Craig and L. Gauvin. 2007.
"Physical Activity, Cardiorespiratory Fitness and Body Mass Index
as Predictors of Substantial Weight Gain and Obesity." Canadian
Journal of Public Health 98:121-24.
Cairney, J. and T.J. Wade. 1998. "Correlates of Body Weight in
the 1994 Population Health Survey." International Journal of
Obesity 22:584-91.
Carr, D. and M.A. Friedman. 2005. "Is Obesity Stigmatizing?
Body Weight, Perceived Discrimination, and Psychological Well-Being in
the United States." Journal of Health and Social Behaviour
46:244-59.
Chandola, T. and C. Jenkinson. 2000. "The New UK National
Statistics Socio-Economic Classification (NS-SEC): Investigating Social
Class Differences in Self-Reported Health Status." Journal of
Public Health Medicine 22:182-90.
Clarke, L.H. 2001. "Older Women's Bodies and the Self:
The Construction of Identity in Later Life." The Canadian Review of
Sociology and Anthropology 38:441-64.
Commission on the Social Determinants of Health. 2008. Closing the
Gap in a Generation: Health Equity through Action on the Social
Determinants of Health. Geneva: World Health Organisation.
Feinstein, J. 1993. "The Relationship between Socioeconomic
Status and Health: A Review of the Literature." Milbank Quarterly
71:279-322.
Frohlich, K.L, N. Ross and C. Richmond. 2006. "Health
Disparities in Canada Today: Some Evidence and a Theoretical
Framework." Health Policy 79:132-43.
Frost, L. 2001. Young Women and the Body: A Feminist Sociology. New
York: Palgrave.
Goffman, E. 1959. The Presentation of Self in Everyday Life. New
York: Anchor Books.
Grogan, S. 1999. Body Image: Understanding Body Dissatisfaction in
Men, Women, and Children. New York: Routledge.
Hall, K.D., A.M. Stephen, B.A. Reeder, N. Muhajarine and G. Lasiuk.
2003. "Diet, Obesity and Education in Three Age Groups of
Saskatchewan Women." Canadian Journal of Dietetic Practice
64:181-86.
House, J.S. 2001. "Understanding Social Factors and
Inequalities in Health: 20th Century Progress and 21st Century
Prospects." Journal of Health and Social Behavior 43:125-42.
Humphries, K.H. and E. van Doorslaer. 2000. "Income-Related
Health Inequalities in Canada." Social Science and Medicine
50:663-71.
Janssen, I., W.F. Boyce, K. Simpson and W. Pickett. 2006.
"Influence of Individual- and AreaLevel Measures of Socioeconomic
Status on Obesity, Unhealthy Eating, and Physical Inactivity in Canadian
Adolescents." American Journal of Clinical Nutrition 83:139-45.
John, U., M. Hanke, J. Grothues and J.R. Thyrian. 2006.
"Validity of Overweight and Obesity in a Nation Based on
Self-Report Versus Measurement Device Data." European Journal of
Clinical Nutrition 60:372-77.
Katzmarzyk, P.T. 2002. "The Canadian Obesity Epidemic: An
Historical Perspective." Obesity Research 10:666-74.
Klarenbach, S., R. Padwal, A. Chuck and P. Jacobs. 2006.
"Population-Based Analysis of Obesity and Workforce
Participation." Obesity 14:920-27.
Krieger, N., J.T. Chen and J.V. Selby. 1999. "Comparing
Individual-Based and HouseholdBased Measures of Social Class to Assess
Class Inequalities in Women's Health: A Methodological Study of 684
US Women." Journal of Epidemiology and Community Health 53:612-23.
Krieger, N., D.R. Williams and N.E. Moss. 1997. "Measuring
Social Class in US Public Health Research: Concepts, Methodologies, and
Guidelines." Annual Review of Public Health 18:341-78.
Kuhle, S. and P.J. Veugelers. 2008. "Why Does the Social
Gradient in Health Not Apply to Overweight?" Health Reports
19:7-15.
Lantz, P.M., J.S. House, R.P. Mero and D.R. Williams. 2005.
"Stress, Life Events, and Socioeconomic Disparities in Health:
Results from the Americans' Changing Lives Study." Journal of
Health and Social Behavior 46:274-88.
Link, B.G. and J. Phelan. 1995. "Social Conditions as
Fundamental Causes of Disease." Journal of Health and Social
Behavior 36:80-94.
Luciano, L. 2001. Looking Good: Male Body Image in Modern America.
New York: Hill and Wang.
Macintyre, S. and K. Hunt. 1997. "Socio-Economic Position,
Gender and Health: How Do They Interact?" Journal of Health
Psychology 2:315-34.
Mackenbach, J.P., I. Stirbu, A.J.R. Roskam, M.M. Schaap, G.
Menvielle, M. Leinsalu and A.E. Kunst. 2008. "Socioeconomic
Inequalities in Health in 22 European Countries." New Englund
Journal of Medicine 358:2468-81.
Matheson, F.I., R. Moineddin and R.H. Glazier. 2008. "The
Weight of Place: A Multilevel Analysis of Gender, Neighborhood Material
Deprivation, and Body Mass Index among Canadian Adults." Social
Science and Medicine 66:675-90.
McLaren, L. 2007. "Socioeconomic Status and Obesity."
Epidemiologic Reviews 29:29-48.
McLaren, L., C.A. Beck, S.B. Patten, G.H. Fick and C.E. Adair.
2008. "The Relationship between Body Mass Index and Mental Health:
A Population-Based Study of the Effects of the Definition of Mental
Health." Social Psychiatry and Psychiatric Epidemiology 43:63-71.
McLaren, L. and J. Godley. 2009. "Social Class and Body Mass
Index among Canadian Adults: A Focus on Occupational Prestige."
Obesity 17:290-99.
McVey, G., S. Tweed and E. Blackmore. 2005. "Correlates of
Weight Loss and Muscle Gaining Behavior in 10- to 14-Year-Old Males and
Females." Preventive Medicine 40: 1-9.
Mulatu, M.S. and C. Schooler. 2002. "Causal Connections
between Socio-Economic Status and Health: Reciprocal Effects and
Mediating Mechanisms." Journal of Health and Social Behavior
43:22-41.
Orbach, S. 1979. Fat is a Feminist Issue. New York: Berkley.
Orbach, S. 1986. Hunger Strike: The Anorectic's Struggle as a
Metaphor for Our Age. London: Faber and Faber.
Ostbye, T., J. Pomerleau, M. Speechley, L.L. Pederson and K.N.
Speechley. 1995. "Correlates of Body Mass Index in the 1990 Ontario
Health Survey." Canadian Medical Association Journal 152:1811-17.
Peralta, R.L. 2003. "Thinking Sociologically about Sources of
Obesity in the United States." Gender Issues 21:5-16.
Pierard, E., N. Buckley and J. Chowhan. 2004. "Bootstrapping
Made Easy: A STATA ADO File." The Research Data Centres Technical
Bulletin (Statistics Canada Catalogue No. 12-002-XIE) 1:20-36.
Pipher, M. 1995. Hunger Palos: The Modern Woman's Tragic
Quest/br Thinness. New York: Ballantine Books.
Pope, H.G., K.A. Phillips and R. Olivardia. 2000. The Adonis
Complex: The Secret Crisis of Male Body Obsession. New York: The Free
Press.
Poudrier, J. 2007. "The Geneticization of Aboriginal Diabetes
and Obesity: Adding Another Scene to the Story of the Thrifty
Gene." The Canadian Review of Sociology and Anthropology 44:237-61.
Power, E.M. 1999. "An Introduction to Pierre Bourdieu's
Key Theoretical Concepts." Journal for the Study of Food and
Society 3:48-52.
Power, E.M. 2005. "Determinants of Healthy Eating among
Low-Income Canadians." Canadian Journal of Public Health 96(Suppl
3):S37-43.
Price, J. and M. Shildrick. 1999. Feminist Theory and the Body: A
Reader. New York: Routledge.
Prus, S.G. 2007. "Age, SES, and Health: A Population Level
Analysis of Health Inequalities Over the Lifecourse." Sociology of
Health and Illness 29:275-96.
Rahkonen, O., M. Laaksonen, P. Martikainen, E. Roos and E. Lahelma.
2006. "Job Control, Job Demands or Social Class? The Impact of
Working Conditions on the Relation between Social Class and
Health." Journal of Epidemiology and Community Health 60: 50-54.
Raphael, D., R. Labonte, R. Colman, K. Hayward, R. Torgerson and J.
Macdonald. 2006. "Income and Health in Canada: Research Gaps and
Future Opportunities." Canadian Journal of Public Health 97(Suppl
3):S16-23.
Rashad, I. 2003. "Assessing the Underlying Economic Causes and
Consequences of Obesity." Gender Issues 21:17-29.
Ricciuto, L.E. and V.S Tarasuk. 2007. "An Examination of
Income-Related Disparities in the Nutritional Quality of Food Selections
among Canadian Households from 1986-2001." Social Science and
Medicine 64:186-98.
Rose, G. 1992. The Strategy of Preventive Medicine. Oxford, UK:
Oxford University Press.
Ross, C.E. and C.-L. Wu. 1995. "The Links between Education
and Health." American Sociological Review 60:719-45.
Ross, N., D. Crouse, S. Tremblay, S. Khan, M. Tremblay and J.-M.
Berthelot. 2007. "Body Mass Index in Urban Canada: Neighborhood and
Metropolitan Area Effects." American Journal of Public Health
97:500-508.
Sanchez-Vaznaugh, E.V., I. Kawachi, S.V. Subramanian, B.N. Sanchez
and D. Acevedo-Garcia. 2009. "Do Socioeconomic Gradients in Body
Mass Index Vary by Race/Ethnicity, Gender, and Birthplace?"
American Journal of Epidemiology 169:1102-12.
Scambler, G. and P. Higgs. 1999. "Stratification, Class and
Health: Class Relations and Health Inequalities in High Modernity."
Sociology 33:275-96.
Shields, M. and M. Tjepkema. 2006. "Trends in Adult
Obesity." Health Reports (Statistics Canada, Catalogue 82-003)
17:53-67.
Sobal, J. and A.J Stunkard. 1989. "Socioeconomic Status and
Obesity: A Review of the Literature." Psychological Bulletin
105:260-75.
Spencer, E.A., P.N. Appleby, G.K. Davey and T.J. Key. 2002.
"Validity of Self-Reported Height and Weight in 4808 EPIC-Oxford
Participants." Public Health Nutrition 54:561-65.
Statistics Canada. 2005. Canadian Community Health Survey 2003:
User Guide for the Public Use Microdata File. Ottawa: Statistics Canada.
Swami, V., A. Furnham, R. Amin, J. Chaudri, K. Joshi, S. Jundi, R.
Miller, J. Mirza-Begum, F.N. Begum, P. Sheth and M.J Tovee. 2008.
"Lonelier, Lazier, and Teased: The Stigmatizing Effect of Body
Size." The Journal of Social Psychology 148:577-93.
Tabachnick, B.G. and L.S. Fidell. 1996. Using Multivariate
Statistics. New York: Harper Collins College Publishers.
Tjepkema, M. 2006. "Adult Obesity." Health Reports
(Statistics Canada, Catalogue 82-003) 17: 9-25.
Townsend, P., N. Davidson and M. Whitehead, eds. 1992. Inequalities
in Health: The Black Report and the Health Divide. Harmondsworth, UK:
Penguin Books.
Tremblay, M.S., P.T. Katzmarzyk and J.D. Williams. 2002.
"Temporal Trends in Overweight and Obesity in Canada,
1981-1996." International Journal of Obesity and Related Metabolic
Disorders 26:538-43.
Tremblay, M.S., C.E. Perez, C.I. Ardern, S.N. Bryan and P.T.
Katzmarzyk. 2005. "Obesity, Overweight and Ethnicity." Health
Reports (Statistics Canada, Catalogue 82-003) 16: 23-34.
Trovato, F. and N. Lalu. 2007. "From Divergence to
Convergence: The Sex Differential in Life Expectancy in Canada,
1971-2000." The Canadian Review of Sociology and Anthropology
44:101-23.
Tugwell, P. and B. Kristjansson. 2004. "Moving from
Description to Action: Challenges in Researching Socio-Economic
Inequalities in Health." Canadian Journal of Public Health
95:85-86.
Veenstra, G. 2007. "Social Space, Social Class and Bourdieu:
Health Inequalities in British Columbia, Canada." Health and Place
13:14-31.
Ward, H., V. Tarasuk and R. Mendelson. 2007. "Socioeconomic
Patterns of Obesity in Canada: Modeling the Role of Health
Behaviour." Applied Physiology, Nutrition, and Metabolism
32:206-16.
Warin, M., K. Turner, V. Moore and M. Davies. 2008. "Bodies,
Mothers and Identities: Rethinking Obesity and the BMI." Sociology
of Health and Illness 30:97-111.
Willms, J.D., M.S. Tremblay and P.T. Katzmarzyk. 2003.
"Geographic and Demographic Variation in the Prevalence of
Overweight Canadian Children." Obesity Research 11:668-73.
World Health Organization. 1997. Obesity: Preventing and Managing
the Global Epidemic. Geneva: WHO.
Zhang, Q. and Y Wang. 2004. "Socioeconomic Inequality of
Obesity in the United States: Do Gender, Age, and Ethnicity
Matter?" Social Science and Medicine 58:1171-80.
JENNY GODLEY AND LINDSAY MCLAREN University of Calgary
(1) There is also an extensive literature on the reciprocal effect
of health on SES (e.g., see Mulatu and Schooler 2002), and in particular
the possible reciprocal effect of BMI/obesity on SES. While the effect
of weight on SES is not the focus of this paper, we acknowledge the
difficulty of making causal arguments regarding health and SES,
especially when relying on cross-sectional data.
(2.) For example, eating less healthy food could be considered a
material mechanism (fruits and vegetables tend to be more expensive than
less healthy food) or a cultural mechanism (eating "health
foods" is more valued by those with higher education). Many
lifestyle and behavioral factors could be considered both.
(3.) With apologies to the sociobiologists, and pleading data
limitations, we leave aside the issue of genetics in this paper.
(4.) This decision eliminated just under 4,000 respondents,
approximately 87 percent of whom were female, and 99 percent of whom
scored above 3.29 on the standardized BMI score.
(5.) We also ran all models using total household income divided by
total household size and divided into quartiles. The results were
substantively the same.
Jenny Oodley, Department of Sociology, University of Calgary, 2500
University Dr. NW, Calgary, AB, Canada T2N 1N4. E-mail:
[email protected]
Table 1
Distribution of Independent Variables among Men and Women of Working
Age (25 to 64) in the Canadian Community Health Survey, Cycle 2.1
Valid % or mean (standard
deviation [SD])
Men Women
Socioeconomic status variables
Household income, (a) N 30,699 30,232
1st quartile (lowest) 19.48% 24.92%
2nd quartile 25.88% 27.93%
3rd quartile 25.76% 24.56%
4th quartile (highest) 28.88% 22.60%
Education, N 36,797 39,340
Less than high school 14.54% 13.95%
High school graduate 24.52% 26.53%
Some postsecondary 36.62% 36.35%
Bachelor's degree or higher 24.32% 23.16%
Sociodemographic variables
Age, N 37,578 39,921
M = 43.47 M = 43.76
(SD = 10.68) (SD = 10.64)
Marital status, N 37,508 39,827
Married/common law 74.35% 73.03%
Separated/divorced 7.03% 10.76%
Single 18.02% 13.65%
Widowed 0.60% 2.57%
Cultural/racial origin, N 36,572 39,000
White 83.73% 84.24%
Black 1.66% 1.78%
East Asian 5.74% 6.22%
South Asian 4.43% 3.04%
Aboriginal 0.96% 1.16%
Other 3.47% 3.55%
Lifestyle variables
Physical activity 15 minutes 36,837 39,679
or more per month, N M = 23.95 M = 24.79
(SD = 23.16) (SD = 22.77)
Daily fruit and vegetable 35,750 38,795
consumption, N M = 3.96 M = 4.82
(SD = 2.29) (SD = 2.51)
Type of smoker, N 37,386 39,779
Daily 22.91% 19.25%
Occasionally 5.66% 4.73%
Not at all 71.43% 76.02%
Frequency of drinking alcohol, N 31,843 31,539
Less than once a month 13.55% 27.39%
1-3 times a month 21.54% 27.58%
1-3 times a week 45.75% 35.87%
4+ times a week 19.16% 9.16%
Work hours, N 37,578 39,921
Do not work for pay 14.11% 25.68%
Work part-time (1-29 hours/ 13.88% 27.98%
week)
Work full-time (30-40 hours/ 35.31% 32.45%
week)
Work slightly over full-time 21.97% 9.75%
(41-50 hours/week)
Work greatly over full-time 14.72% 4.14%
(51+ hours/week)
(a) Total household income divided by the square root of household
size. Note that quartiles were created with the data for men and
women combined.
Table 2
Regression of BMI on Income, Sociodemographic Controls, and Lifestyle
Variables
Men
Model A Model B Model C
Household income
1st quartile (low) -.615 ** -.307 * -.218
(.111) (.108) (.115)
2nd quartile -.133 -.038 .003
(.093) (.089) (.090)
3rd quartile .060 .112 .138
(.094) (.093) (.091)
4th quartile (high) Reference
Age .023 ** .027 **
(.004) (.004)
Marital status
Divorced/separated -.616 ** -.502 **
(.115) (.112)
Single -.704 ** -.597 **
(.103) (.104)
Widowed .229 .249
(.321) (-.320)
Married/common law Reference
Cultural/racial origin
Black -.430 -.568
(.319) (.313)
East Asian -2.948 ** -3.149 **
(.174) (-.180)
South Asian -1.246 ** -1.385 **
(.300) (.307)
Aboriginal 1.010 1.030
(.429) (.410)
Other -.454 -.431
(.250) (.241)
White Reference
Fruit and vegetables -.076 **
(.016)
Exercise -.003
(.002)
Smoking
Daily -.929 **
(.100)
Occasionally -.208
(.154)
Do not smoke Reference
Alcohol
1-3 times a month -.099
(.136)
1-3 times a week -.257
(.123)
4+ times a week -1.079 **
(.131)
Less than once a month Reference
Hours worked
No paid work .090
(.140)
Work 41-50 hours -.129
(.109)
Work part-time per week .275 *
(.088)
Work 51+ hours/week .481 **
(.106)
Work full-time Reference
Constant 26.678 ** 25.960 ** 26.525 **
(.063) (.175) (.239)
N 25,804 25,804 25,804
[R.sup.2] .004 .044 .066
Women
Model A Model B Model C
Household income
1st quartile (low) 1.065 ** 1.271 ** .878 **
(.134) (.135) (.159)
2nd quartile .959 ** 1.052 ** .745 **
(.125) (.118) (.126)
3rd quartile .520 ** .635 ** .426 **
(.123) (.119) (.122)
4th quartile (high)
Age .071 ** .077 **
(.004) (.004)
Marital status
Divorced/separated -.399 -.328
(.167) (.163)
Single -.012 .082
(.137) (.139)
Widowed -.547 -.469
(.324) (.308)
Married/common law
Cultural/racial origin
Black .274 -.129
(.473) (.482)
East Asian -2.785 ** -3.284 **
(.200) (.207)
South Asian -1.247 ** -1.547 **
(.354) (.346)
Aboriginal 1.645 ** 1.687 **
(.335) (.325)
Other -.226 -.511
(.256) (.262)
White
Fruit and vegetables -.051 *
(.018)
Exercise -.012 **
(.002)
Smoking
Daily -.658 **
(.120)
Occasionally -.641 **
(.155)
Do not smoke
Alcohol
1-3 times a month -.614 **
(.130)
1-3 times a week -1.597 **
(.119)
4+ times a week -2.137 **
(.150)
Less than once a month
Hours worked
No paid work -.329
(.154)
Work 41-50 hours -.066
(.120)
Work part-time per week -.020
(.179)
Work 51+ hours/week -.012
(.203)
Work full-time
Constant 24.295 ** 21.338 ** 23.057 **
(.083) (.200) (.257)
N 24,291 24,291 24,291
[R.sup.2] .008 .050 .083
Note: Unstandardized regression coefficients and (standard errors)
shown.
* p <.01.
** p < .001.
Model A, unadjusted; Model B, adjusted for sociodemographic controls;
Model C, adjusted for sociodemographic controls and lifestyle
variables; BMI, body mass index.
Table 3
Regression of BMI on Education, Sociodemographic Controls, and
Lifestyle Variables
Men
Model A Model B Model C
Education
Less than high school 1.323 ** 1.066 ** 1.294 **
(.124) (.123) (.124)
High school graduate 1.054 ** .944 ** 1.057 **
(.095) (.092) (.093)
Some postsecondary 1.096 ** .882 ** .960 **
(.079) (.079) (.079)
BA or higher Reference
Age .020 ** .025 **
(.003) (.003)
Marital status
Divorced/separated Reference -.731 ** -.574 **
(.106) (.104)
Single
Widowed -.776 ** -.618 **
(.091) (.093)
Married/common law .507 .604
(.375) (.386)
Cultural/racial origin
Black -.605 -.757
(.325) (.327)
East Asian -2.926 ** -3.076 **
(.158) (.162)
South Asian -1.099 ** -1.243 **
(.255) (.261)
Aboriginal .794 .907
(.390) (.372)
Other -.412 -.377
White Reference (.224) (.214)
Fruit and vegetables -.054 **
(.015)
Exercise -.002
(.001)
Smoking
Daily -1.168 **
(.092)
Occasionally Reference -.260
(.147)
Do not smoke
Alcohol Reference
1-3 times a month .005
(.124)
1-3 times a week -.185
(.108)
4+ times a week -.850 **
(.122)
Less than once a month Reference
Hours worked
No paid work -.166
(.120)
Work part-time -.137
(.097)
Work 41-50 hours/week .259 *
(.082)
Work 51+ hours/week .420 **
(.098)
Work full-time
Constant 25.696 ** 25.355 ** 25.729 **
(.061) (.161) (.213)
N 29,697 29,697 29,697
[R.sup.2] .016 .058 .082
Women
Model A Model B Model C
Education
Less than high school 2.283 ** 1.757 ** 1.535 **
(.158) (.161) (.167)
High school graduate 1.414 ** 1.200 ** 1.102 **
(.115) (.112) (.115)
Some postsecondary 1.183 ** 1.023 ** .940 **
(.106) (.102) (.103)
BA or higher
Age .059 ** .065 **
(.004) (.004)
Marital status
Divorced/separated -.218 -.188
(.141) (.138)
Single
Widowed .177 .263
(.127) (.127)
Married/common law -.369 -.337
(.278) (.265)
Cultural/racial origin
Black .173 -.297
(.403) (.410)
East Asian -2.597 ** -3.183 **
(.177) (.191)
South Asian -1.307 ** -1.650 **
(.331) (.332)
Aboriginal 2.111 ** 2.145 **
(.368) (.350)
Other .223 -.024
White (.265) (.268)
Fruit and vegetables -.045 *
(.016)
Exercise -.013 **
(.002)
Smoking
Daily -.891 **
(.110)
Occasionally -.825 **
(.151)
Do not smoke
Alcohol
1-3 times a month -.598 **
(.111)
1-3 times a week - 1.515
(.105)
4+ times a week -2.021 **
(.140)
Less than once a month
Hours worked
No paid work -.262
(.121)
Work part-time -.007
(.105)
Work 41-50 hours/week .025
(.162)
Work 51+ hours/week .081
(.172)
Work full-time
Constant 23.826 ** 21.535 ** 23.100 **
(.084) (.191) (.232)
N 30,224 30,224 30,224
[R.sup.2] .023 .056 .090
Note: Unstandardized regression coefficients and (standard errors)
shown.
* p < .01.
** p < .001.
Model A, unadjusted; Model B, adjusted for sociodemographic controls;
Model C, adjusted for sociodemographic controls and lifestyle
variables; BMI, body mass index.
