1. Social Inequalities and Exercise during Adulthood: Toward an Ecological Perspective
Author(s): Joseph G. Grzywacz and Nadine F. Marks
Source: Journal of Health and Social Behavior, Vol. 42, No. 2 (Jun., 2001), pp. 202-220
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/3090178
Accessed: 13/05/2009 19:06
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Health and Social Behavior.
http://www.jstor.org

2. Social Inequalities and Exercise during Adulthood: Toward an
Ecological Perspective*
JOSEPH G. GRZYWACZ
Universityof NorthernIowa
NADINE F. MARKS
Universityof Wisconsin-Madison
2001,Vol42 (June):
Journal Health SocialBehavior
of 202-220
and
Grounded in ecological theory, this study examines the association among par-
ticipation in regular vigorous exercise and social status, aspects of prominent
life settings, interactions between life settings, and more proximal individual
resources and processes using data from the National Survey of Midlife
Development in the United States (N = 3,032). Among women, a higher level of
earnings was associated with more vigorous exercise, yet those women with
more education had a steeper decline in exercise across adulthood. Among men,
those with the lowest level of education had the steepest decline in physical
activity across adulthood, and earnings did not affect exercise patterns. Less
participation in vigorous exercise among blacks, in contrast to nonblacks, was
explained by their tendency to live in less safe neighborhoods and having more
functional health problems. Finally, contextual factors from multiple domains
were independently associated with participation in regular exercise.
Consistent with ecological theory, these results suggest that interventions to
promote exercise habits among adults need to consider the independent and
interactive effects of multiple contextualfactors.
Regular physical activity has been demon- ic and acute morbidity (Breslow and Breslow
strated to promote longevity (Berkman and 1993; U.S. Departmentof Health and Human
Breslow 1983; Fried et al. 1998; Paffenbarger Services 1996), and facilitate higher levels of
et al. 1994; U.S. Department of Health and psychological well-being throughout adult-
Human Services 1996), reduce rates of chron- hood (Albert 1995; Fries and Crapo 1985;
Folkins and Sime 1981; Rowe and Kahn 1987;
Tkachukand Martin1999; U.S. Department of
*Please address correspondence Joseph G. Health and Human Services 1996). Most
to:
Divisionof Health Schoolof adultsremaininactive despite widespreadpro-
Promotion,
Grzywacz,
and
Health,PhysicalEducation, LeisureServices,
gramming to initiate regularphysical activity
University of Northern Iowa, 203 Wellness/
Cedar Falls, IA 50614-0241. (Public Health Service 1991). National esti-
Recreation Center,
mates indicate that 30 percent of American
joe.grzywacz@uni.edu:
This research was supportedby a National adults remain completely sedentary,and only
Traineeship 14-23 percent of adults engage in enough
Institute Mental
of Health Post-Doctoral
(MH19958), the National Institute on Aging physical activity to achieve health-related ben-
the
(AG12731), National Institute Mental
of Health efits (Centers for Disease Control 1997; U.S.
(MH61083),and the John D. and Catherine T. of Health and Human Services
MacArthurFoundationResearch Network on Department
1996).
Successful MidlifeDevelopment. significant
A por-
occurred Individuals in lower social status positions
tion of the preparation this manuscript
of
while the lead author was affiliated with the participatein less physical activity than their
of higher status counterparts. Older adults,
of and
Department Psychology SocialBehavior
theUniversity California,
of Irvine. women, blacks, and individuals of lower
202

3. AND 203
INEQUALITIESEXERCISE
SOCIAL
socioeconomic status are consistentlyfound to AND EMPIRICAL
THEORETICAL
BACKGROUND
exercise less regularlyand are more likely to
be completely sedentarythan younger adults,
Ecological theory, as explicated by
men, whites, and higher socioeconomic status
Bronfenbrenner and colleagues (Bronfen-
individuals (for reviews see Blair 1988;
brenner1979; Bronfenbrenner Ceci 1994;
and
Centers for Disease Control 1997; Dishman
Bronfenbrennerand Morris 1998), can serve
1991; King et al. 1992; U.S. Department of
as a valuable tool for further understanding
Health and Human Services 1996). Given this
and modeling the determinants of positive
evidence, age, gender, race, and educationare health behaviors (Fitzgerald et al. 1994;
typically included as important covariates
McLeroy et al. 1988; Sallis and Owen 1997;
when developing and testing models of physi- Stokols 1996). Bronfenbrenner's model posits
cal activity. However, inattentively quot;control- that behaviors such as physical activity are
ling forquot; social status is a major limitation in influenced by a variety of factors from multi-
health research (Marmot, Kogevinas, and ple ecological levels (e.g., individual,
Elston 1987). Instead,Marmotand colleagues microsystemic, mesosystemic, and macrosys-
(1987) as well as others (Krieger, Williams, temic) and change as a function of develop-
and Moss 1997; Williams 1990) argue that mental and historical time. Consequently,
more researchdirectlyexaminingthe effects of using an ecological perspective, scholars can
integratea rich arrayof known determinants of
social status on health-relatedphenomena is
physical activity (for reviews see Blair 1988;
critically needed since social status creates a
Dishman 1991; King et al. 1992; U.S.
broad context that shapes routine experiences
Department of Health and Human Services
that affect health-related behavior.
1996) into models of exercise to better specify
In this study, we seek to expand our under-
predictive models for different periods in the
standing of physical activity patternsby con-
life course. In addition to being more con-
sidering exercise behavior within the broad ducive to a more comprehensive theory of
context of social status, and examining how
health behavior, an ecological approach also
contextualfactors from prominentlife settings
has severalpracticalbenefits. Most notably,an
and more proximal individual resources and
ecological approachattenuatesthe possibility
processes help account for and uniquely con- of quot;victim-blamingquot;(e.g., Becker 1986) by
tributeto differencesin exercise. Guidedby an
drawing attention to additional targets for
ecological perspective and previous research, intervention beyond the individual. When
we theorize that the often-noted declines in acted upon and changed, these socio-environ-
activity by age, and the frequentlycited differ- mental factors frequently bring about more
ences in physical activity by gender,race, and sustainable change in individual and popula-
tion health behavior(Green and Krueter1999;
socioeconomic status do not act independently
Stokols, Pelletier,and Fielding 1996).
of each other.Thus, in this study we systemat-
ically examine the independentand interactive
effects of age, gender,andmultipledimensions
Social Status and Physical Activity
of socioeconomic status on physical activity
patterns throughout adulthood. Consistent
Severalpersonalcharacteristics representing
with previousscholars,we furthertheorizethat
individuals' relative location in status hierar-
social status partially allocates contextual
chies predict participationin regularphysical
resources, opportunities, and barriers at the
activity. Lower socioeconomic status, older
individual,family,work,and communitylevels
age, being female, and being black are consis-
that may influence physical activity patterns.
tently associatedwith less participationin reg-
Therefore,we examine the independenteffects ular physical activity (for reviews see Blair
of individual-level, family- and work-related, 1988; Dishman 1991; King et al. 1992; U.S.
as well as community experiences on regular Department of Health and Human Services
exercise, and we consider the extent to which 1996). Like previous scholars (Ross and Wu
the effects of social status operate through 1995; Stronegger,Freidl, and Rasky 1997), we
theorize that social location in status hierar-
shaping differences in these everyday experi-
chies is an importantconditioning factor for
ences.

4. 204 JOURNALOF HEALTHAND SOCIALBEHAVIOR
the allocation of resources, opportunities,and earnings on exercise habits will be more pro-
constraintsthat influence behaviors related to nounced among men than women.
health, such as exercise. Time is a centralconcept in ecological mod-
Socioeconomic status is consistentlyassoci- els (Bronfenbrenner1995), and it draws our
ated with health behaviorsand health behavior attentionto the enduringcontext of socioeco-
change: Lower socioeconomic status individu- nomic status for the developmentand mainte-
als participatein fewer positive health behav- nance of positive health behaviors such as
iors (e.g., exercise, maintaininghealthy body exercise. The quot;cumulativeadvantagequot;hypoth-
weight) and change their negative health esis (Ross and Wu 1996) directlyaddressesthe
behaviors (e.g., smoking) at a slower rate than issue of time by positing that the longer an
higher socioeconomic status individuals individual is exposed to a protective factor,
(Berkman and Breslow 1983; Blaxter 1990; such socioeconomic advantage,the greaterthe
National Center for Health Statistics 1998; health-related benefit. Scholars have found
Piani and Schoenborn 1993; Ross and Wu some supportfor the cumulativehealth status
1995; Stroneggeret al. 1997; U.S. Department benefits of education(e.g., House et al. 1990;
of Health and Human Services 1996). House et al. 1994; Ross and Wu 1996), howev-
Evidence from different studies however sug- er this question has not yet been considered
gests that differentindicatorsof socioeconom- with respectto participation regularphysical
in
ic status (e.g., education and earnings) may activity.Assuming that socioeconomic advan-
have differentassociations with physical activ- tage provides cumulative benefits for contin-
ity, and that these associations may differ by ued participation regularexercise over time,
in
gender (Cauley et al. 1991; Fordet al. 1991) or we hypothesizethat a wider gap in the benefits
measure of physical activity (Cauley et al. of educationand earningswill be more appar-
ent at older ages than younger ages.
1991).
Three importantfeatures are demonstrated
by the previous studies that supportthe mean-
ingfulness of an ecological model in the study Contextualand IndividualCorrelatesof
of physical activity. First, socioeconomic sta- Physical Activity
tus is best conceptualized as being multidi-
mensional, and each independent dimension Certainrecurring patternsin everydaysocial
(e.g., education and income) has the ability to life exert a disproportionate amount of influ-
influence physical activity habits in different ence on well-being (Stokols 1996) and, pre-
ways. For example, educationmight influence sumably,on behaviors related to health. Most
health behaviors though a greater ability to adults are nested within families, occupations,
receive and interpretinformationregardingthe and communities; consequently, comprehen-
health benefits of regular exercise. Earnings, sive models of physical activity need to con-
on the other hand, independent of education, sider social and environmentalfactors from
each of these domains or settings
might promote physical activity throughaddi-
tional discretionaryincome that allows pursu- (Bronfenbrenner 1979). Unfortunately,most
ing physically active hobbies during leisure health behavior research does not give ade-
time. Next, this researchhighlights the impor- quateattentionto notablecontextualor ecolog-
tance of recognizing the specific type of phys- ical factorsthatmay supportor underminepar-
ical activity being studied (e.g., amount of ticipationin regularphysical activity.However,
leisure time physical activity per week versus researcherscan use middle-rangetheory and
kilocalories expended per day). Finally,multi- empirical evidence from different fields (e.g.,
ple strands of evidence suggest that the link family studies, occupational health, urban
between socioeconomic status and health is planning) to identify high-powered leverage
gendered whereby education and earnings points (i.e., factors that exert, or potentially
have consistent health effects among men but exert, a disproportionate amount of influence;
not women (see Marmot et al. 1987; see Stokols 1996) for influencing exercise.
McDonough et al. 1999). We hypothesize that Thefamily microsystem. The family is a pri-
a higher level of education and a higher level mary life domain for most adults. Since indi-
of earnings will be independentlyassociated viduals in specific family roles, such as spouse
with more regular physical activity. We also or parent,are frequentlyfound to participatein
hypothesize that the effects of education and fewer health-risk behaviors (Chilcoat and

5. 205
INEQUALITIES EXERCISE
AND
SOCIAL
Breslau 1996; Horwitz and Raskin White Cooper 1993; Frone, Russell, and Cooper
1991; Umberson 1987), the family is believed 1997), suggesting that incompatiblework and
to be an importantsource of social control of family responsibilitiesmay underminepositive
behaviors related to health (Umberson 1987). health behaviors such as exercise (Backett
Similarly, family life provides an important 1992; Backett and Davison 1995). However,
general context for health promotingbehavior recent theory emphasizes the quality of fit
(Doherty and McCubbin 1985; Pratt 1976; between work and family as opposed to expe-
Walsh 1993). Previous research, for example, riences of strainsalone (Barnett1998). That is,
has demonstrated that emotionally close fami- both the benefits (Sieber 1974) and the strains
ly relationshipsand happy,stable, and satisfy- (Goode 1960) of role accumulationneed to be
ing marriagespredictless healthrisk behaviors consideredwhen examiningthe health impacts
among adults (Doherty and Harkaway 1990; of the work-family interface. Based upon
Franks, Campbell, and Shields 1992; Pratt recent work-family theory (Barnett 1999;
1976; Wickrama et al. 1997a). Following Grzywacz and Marks 2000), we hypothesize
that more positive spilloverbetween work and
social controltheory,we hypothesizethatmar-
ried individualsand parentswill exercise more family and less negative spillover between
frequentlythan single or childless individuals. work and family will be associated with more
Consistent with family systems theory, we regularphysical activity.
hypothesize that individualsin an emotionally The researchcited here shares one common
limitation. Assuming that specific health
close family and an emotionally close mar-
behaviors are reflections of a common latent
riage will exercise more.
The work microsystem.The workplace is a quot;lifestylequot; construct,researcherstypically use
target for implementing health promotion summed indices of health-related behavior
(e.g., non-smoking, non-problemalcohol use,
interventions(PublicHealth Service 1991), yet
and regular exercise) or substance use as the
we lack a body of clearly defined research
dependent variable. Consequently,it remains
explicitly examiningthe impact and the mech-
unclear what effect, if any, family, work, and
anisms through which the workplace influ-
work-familyspilloverwill have specifically on
ences behavior related to health (for recent
physical activity,since wellness enhancingand
review see Eakin 1997). Nonetheless, a variety
risk-taking behaviors (e.g., regular physical
of job characteristics,particularlyhigh deci-
activity versus substance use; Vickers,
sion latitude,have been found to predict fewer
Conway, and Hervig 1990) are qualitatively
health-risk behaviors (House et al. 1986;
different(for review, see Gochman 1997), and
Mensch and Kandel 1988; Weidneret al. 1997;
may not share the same contextual determi-
Wickramaet al. 1997b). Based upon occupa-
tional stress theory (House 1981; Karasekand nants.
Community.A wide range of community
Theorell 1990) and previous research, we
characteristicsmay influence an individual's
hypothesize that having more decision latitude
ability to practice various behaviorsrelated to
on the job will be associatedwith more regular
health (Cheadle et al. 1999; Taylor, Repetti,
physical activity.We also hypothesizethat due
and Seeman 1997). Evidence from different
to time constraints,a greaternumberof hours
sources indicates that individuals who live in
worked per week will be associated with less
unsafe communitiesor neighborhoodsare less
exercise.
likely to engage in regular physical activity
The work-family mesosystem. Ecological
(Centersfor Disease Control 1999; Eyler et al.
theory also contends that unique contributions
1998; Ross 1993; Sallis et al. 1997). Although
to the overall context of human development
some reportssuggest that neighborhoodsafety
are derived from interactionsbetween two or
serves as a physical barrierto regularactivity
more life settings (i.e., mesosystems;
throughfear of victimization (e.g., Centersfor
1979). Managingand integrat-
Bronfenbrenner
Disease Control 1999; Ross 1993), the mecha-
ing work and family is an increasinglysignifi-
nisms linking neighborhoodsafety to individ-
cant task for adults today (Bielby 1992;
ual physical activity have not been adequately
Heckhausen1997; Lachmanand Boone-James
tested or identified. We hypothesizethat living
1997). Work-familystrainhas been foundto be
in neighborhoodscharacterizedas unsafe will
associatedwith more alcohol use among adults
be associatedwith less participationin regular
(Bromet, Dew, and Parkinson 1990; Frone,
Barnes, and Farrell 1994; Frone, Russell, and physical activity.

6. JOURNALOF HEALTHAND SOCIALBEHAVIOR
206
ProximalIndividualResources and Processes Physical Activity:An Ecological Perspective
The ecological model of physical activity
An individual'sphysical and psychological
duringadulthoodthat guided this study is por-
characteristicsare perhaps the most proximal
trayed in Figure 1. Consistent with previous
determinantsof physical activity. In our con-
research, Figure 1 suggests that sociodemo-
ceptualization, we include these factors as
graphic characteristics, including multiple
potential proximal processes conditioned by
aspects of socioeconomic status, have direct
more macrosociological processes, working
effects on rates of participationin regular,vig-
throughthe opportunities,constraints,and rel-
orous exercise. Additionally, our model sug-
ative stressors that are allocated through the
gests that these sociodemographiccharacteris-
social hierarchiesindicated by socioeconomic
tics may also have indirecteffects on physical
status, age, gender, and race (see Williams
exercise by influencing the characteristicsof
1990). individuals' daily contexts (family, work, and
Physiologically, progressive reductions in
community)andthe interactions betweenthese
maximal oxygen uptake(V02 max), cardiovas-
contexts (work-family interface). We also
cular function, and declining muscle mass
expect thatthe qualityof these life settingswill
through adulthood are believed to interfere shape more proximalindividualresources and
with individuals'abilities to participatein reg-
processes-including functionaland breathing
ular or prolonged bouts of physical activity. limitations to physical activity, body mass
However,studies clearly indicate that many of index, environmentalmastery,exercise effica-
the physiological declines associated with
cy (i.e., belief that exercise is a good way to
aging are the result of, not the cause of, promotehealth) and control over health-that,
reduced physical activity (see Goldberg, in turn,influence rates of participation exer-
in
Dengel, and Hagberg 1996). This evidence cise.
suggests that the unfolding of social and psy-
chological dynamics throughout adulthood
cause reductions in physical activity (Backett METHOD
and Davison 1995) that become progressively
Data
reinforcedby subsequentreductionsin physio-
logical function. We hypothesize that a higher
The data used for this study are from the
body-mass index, more functional restrictions
National Survey of Midlife Development in
related to activity, and reduced lung capacity
the United States, collected in 1995 by the
will be associated with less regular vigorous
John D. and Catherine T. MacArthur
exercise.
FoundationResearch Network on Successful
Psychological theories and models that
Midlife Development (N = 3,032; 1,471 men
place primary attention on individual-level
and 1,561 women). The originalpurposeof the
beliefs, attitudes, and values (Glanz, Lewis,
National Survey of Midlife Development was
and Rimer 1997) typically guide health behav-
to examine patterns, predictors, and conse-
ior research.Individualprocesses such as more
quences of midlife developmentin the areasof
self-efficacy (Dzewaltowski 1994; Rosen-
physical health, psychological well-being, and
stock, Strecher,and Becker 1988), a belief that social responsibility.Respondentsare a nation-
exercise is an efficient way to promotehealth,
sample of non-institutional-
ally representative
and a greater sense of control over health
ized persons aged 25-74 who have telephones.
(Strickland 1978) have all been theorized and The samplewas obtainedthroughrandomdigit
subsequentlyfound to influence physical activ- dialing, with an oversamplingof older respon-
ity among adults (King et al. 1992; U.S. dents and men made to guaranteea good dis-
Department of Health and Human Services tributionon the cross-classificationof age and
1996). We drawupon existing value-expectan- gender.Samplingweights correctingfor selec-
cy theoriesto hypothesizethata higher level of tion probabilitiesand non-responseallow this
self-efficacy, a greaterbelief in the healthben- sample to match the composition of the U.S.
efits of exercise, and more perceived control populationon age, sex, race, and education.
over health will be associated with more fre- Respondents first participated in a tele-
quent participationin regularphysical activity. phone interviewlasting approximately min-
40

7. 207
INEQUALITIES EXERCISE
AND
SOCIAL
FIGURE 1. An Ecological Model for Regular Vigorous Physical Activity during Adulthood
utes. The response rate for the telephone ques- seasonal effects (Marmot et al. 1991).
tionnaire was 70 percent. Respondentsto the Specifically, respondents were asked: (1)
telephone survey were then asked to complete quot;Duringthe summer,how often do you engage
two self-administered mailback question- in vigorousphysical activity(for example,run-
naires. The response rate for the mailback ning or lifting heavy objects) long enough to
questionnaire was 86.8 percent of telephone work up a sweat?quot;and (2) quot;Duringthe winter,
respondents.This yielded an overall response how often do you engage in vigorous physical
rate of 60.8 percentfor both partsof the survey activity long enough to work up a sweat?quot;
(for a detailed technical reportregardingfield Response categories included several times a
procedures,response rates, and weighting, see week or more (6), about once a week (5), sev-
http://midmac.med.harvard.edu/research.htmleral times a month (4), aboutonce a month (3),
less than once a month (2), or never (1). The
#tchrpt).
two items were highly correlated (r = .89);
consequently, we used the mean of the two
items to assess regular,vigorous exercise. (See
Measures:Dependent Variable
Table 1 for descriptivestatistics for all analyt-
even a small amountof exercise is ic variables.)
Although
better than being sedentary (e.g., Kunst et al.
1999; Dunn et al. 1999; U. S. Departmentof
Health and Human Services 1996), evidence Measures: IndependentVariables
indicatesthatregularexercise in 60-90 percent
Social status. Age was included as a contin-
of an individual'stargetheartrate is most ben-
eficial to overall health (American College of uous variable in these analyses. Gender (1 =
Sports Medicine 1990). Therefore, we con- female) and race/ethnicity (1 = black) were
structeda continuousoutcome measureof reg- included as dichotomousmeasures. Education
ular vigorous exercise using two previously representsthe numberof years of formal edu-
validated measures (Blair 1984; Kohl et al. cation completed and is coded as an ordinal
1988; Washburn al. 1990) adaptedto include variable as follows: 1 = 1-6 years, 2 = 7-8
et

8. JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
208
TABLE 1. Descriptive statistics for all analysis variables
SD
M Range
Dependent Variable
1.75
4.04 1-6
Regular,Vigorous Exercise
Social Status Characteristics
45.30 13.48 25-74
Age
Gender(female = 1) 56.6%
Race/Ethnicity(black = 1) 11.2%
2.41
6.20 1-12
EducationalAttainment
Household Earnings 0-300,000
36,172
39,254
Family & WorkMicrosystems
ParentalStatus(Child < 18 = 1) 41.1%
MaritalStatus (not married- 1) 31.9%
37.3%
High Spouse Emotional Support
30.4%
Low Spouse Emotional Support
.63
3.41 1-4
FamilyEmotional Support
CurrentlyWorkingFor Pay (Yes = 1) 73.8%
43.52 15.53 1-168
HoursWorkedper Weeka
.87 1-5
3.62
Decision Latitudeat Worka
Spillover between Work Familya
&
.74 1-5
2.62
Negative Workto Family
.84 1-5
2.62
Positive Workto Family
2.11 .68 1-5
Negative Familyto Work
1-5
.84
3.41
Positive Familyto Work
Environment
Community
3.36 .56 1-4
NeighborhoodSafety
IndividualProximalProcesses/Resources
9.44-64.02
5.65
26.80
Body Mass Index
1-4
.77
FunctionalRestrictionsfrom Exercise 1.44
0-3
.57
.18
Shortnessof Breathwhen Active
3-21
3.42
15.87
Environmental Mastery
1-6
.77
PerceivedControlover Health 5.36
0-4
Efficacy of Exercise 3.68 .85
Source: National Surveyof Midlife Developmentin the UnitedStates 1995.
Notes: Estimatesare based on weighted data.
aEstimateobtainedfrom only respondentswho were working for pay when they completed the questionaire.
years, 3 = 9-12 years but no diploma or GED, categories to facilitate inclusion of all respon-
4 = GED, 5 = graduated from high school, 6 = dents:(1) Individualswho were unmarried, (2)
1-2 years of college but no degree yet, 7 = 3 or married individuals with a mean response of
more years of college but no degree yet, 8 = less than4 (i.e., low emotionalsupport;used as
graduatedfrom a two-yearcollege or vocation- the contrastcategory),and (3) marriedindivid-
al school, or associate's degree, 9 = graduated uals with a mean response of 4 (i.e., high emo-
from a four- or five-year college, or bachelor's tional support;note: 56.4% of marriedrespon-
degree, 10 = some graduate school, 11 = mas- dents had a mean of 4 for spouse emotional
ter's degree, and 12 = doctoralor otherprofes- support;consequently,we chose this value for
sional degree. Household earningswere coded maximal variability in the measure). Family
continuouslyin thousandsof dollars. emotional supportmeasuredthe level of emo-
Family Characteristics. Spouse emotional tional supportprovidedto the respondentfrom
support measured the level of emotional sup- his or her other family memberswith a paral-
lel 4-item index (a = .83) that was included in
port providedto the respondentfrom his other
spouse using the mean of a 6-item index (a = the model as a continuous measure. Parental
status (i.e., has a child under 18 years old = 1)
.85) adapted from Schuster, Kessler, and
Aseltine (1990) (e.g., quot;How much does your was also included in our model.
spouse or partner really care about you?quot; Work characteristics.Decision latitude was
quot;Howmuch can you rely on him or her for help assessed by summing responses to five items
if you have a serious problem?quot;). Response measuringthe amountof controlthe individual
categories for the index items were not at all has over his or her work environment and
(1), a little (2), some (3), and a lot (4). Spouse tasks, and the specialization of labor (e.g.,
emotional supportwas then divided into three quot;How often do you have a choice in deciding

9. 209
AND EXERCISE
SOCIAL INEQUALITIES
how you do your tasks at work? How often do egories were not at all (1), some (2), a little (3),
you have a choice in deciding what tasks you and a lot (4). Reduced VO2 max was opera-
do at workquot;;a = .87). We also includeda mea- tionalized using three items adaptedfrom the
sure of the number of hours the respondent Rose Questionnaire for angina (Rose et al.
reportedworking(i.e., sum of responsesto two 1982) assessing if respondents get short of
questions regardinghours worked last week in breath (yes/no) while (1) quot;walkingwith other
primaryand additionaljobs). people your age on level ground;quot; quot;walking
(2)
at your own pace on level ground;quot;and (3)
Work-familyspillover. Negative spillover
quot;washing or dressingquot; (a = .74). Body mass
from worktofamily measuredthe respondent's
perception of the extent to which work inter- index was computed using the Quetlet Index
fered with functioningat home by calculating (i.e., weight/height2).
the mean response to four items (e.g., quot;How Control over health was measured using
often does stress at work make you irritableat three items adaptedfrom the WhitehallSurvey
home?quot;; a = .84). Conversely, positive (Marmot et al. 1991), including 1) quot;Keeping
spillover from work to family assessed the healthy depends on things that I can do;quot; (2)
extent to which the respondent felt that their quot;Thereare certainthings I can do for myself to
work promoted better functioning at home reducethe risk of heartattack;quot; (3) quot;There
and
(e.g., quot;How often do the things you do at work are certainthings I can do for myself to reduce
help you deal with personal and practical the risk of getting cancer.quot; Response categories
issues at home?quot;;a = .74). Negative spillover for the control over health items ranged from
from family to work assessed the extent to strongly disagree (1) to strongly agree (6) (a
which the respondentfelt their family life was =.71).
interferingwith their success on the job (e.g., Environmentalmastery is a component of
quot;How often does stress at home make you irri- psychological well-being that assesses an indi-
table at work?quot;; a = .81). Finally, positive vidual's sense of mastery and competence in
spillover from family to work measured the managing the surroundingenvironment(Ryff
extent to which respondents felt their family 1989). Environmentalmastery was measured
life helped them perform better on the job with three items: (1) quot;In general, I feel I am in
(e.g., quot;How often does talkingwith someone at charge of the situation in which I live;quot; (2)
home help you deal with problems at work?quot;; quot;The demands of everyday life often get me
a = .73). Response categories ranged from down;quot;and (3) quot;I am quite good at managing
never (1) to all the time (5). The work-family the many responsibilities of my daily life.quot;
Response categories for the environmental
spillover items were developed for the
National Survey of Midlife Development in mastery items ranged from strongly disagree
(1) to strongly agree (7) (a = .52). Although
the United States.
the estimated reliability of the environmental
Community. Perceived community safety
mastery scale used in the National Survey of
was measuredwith the mean response to four
Midlife Development in the United States is
items developed for the National Survey of
modest, previous research has indicated that
Midlife Development in the United States: (1)
the three-item measure is strongly correlated
quot;I feel safe being out alone in my neighbor-
with its highly reliable parent measure (Ryff
hood during the daytime;quot; (2) quot;I feel safe
being out alone in my neighborhoodat night;quot; and Keyes 1995).
Exercise efficacy assessed the extent to
(3) quot;I could call on a neighbor for help if I
which respondents believe that exercise is a
needed it;quot; and (4) quot;People in my neighbor-
beneficial and preferred way of promoting
Response categoriesfor
hood trusteach other.quot;
health. Respondentswere asked to respond to
the communitysafety items rangedfrom not at
all (1) to a lot (4) (a = .65). a hypothetical scenario during the telephone
interview quot;Assume for a moment that your
Physiological and psychological resources
doctor said you had a heart condition and said
andprocesses. Functional restrictionsto phys-
you could choose either to have coronary
ical activity was assessed with a 4-item scale
bypass surgeryor exercise at least three times
revised from the Medical Outcomes Study
a week for at least half an hour each time.
(Brazier et al. 1992) asking how much the
Which one do you think you would choose?quot;
respondent's health limited walking several
Respondentswere then asked, quot;How sure are
blocks, walking one block, vigorous activity,
you that this is what you would choose?quot; very
and moderateactivity (a = .92). Response cat-

10. 210 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
pay when they completedthe telephone survey
(1), somewhat (2), or not very sure (3). Being
and self-administered questionnaires;conse-
very sure about choosing exercise was coded 3,
quently, some individuals were legitimately
being somewhat sure about exercise was coded
2, being not very sure about exercise was on quot;decision hours
latitude,quot;
quot;missingquot;
worked/week, and each of the work-family
coded 1, and choosing bypass surgery was
spillover measures. Rather than limiting the
coded 0.
sample to employedadults only, we includeda
quot;missing flagquot; indicator variable for each
measurein the model to generate
work-related
Analytic Sequence
more reliable population parameterestimates
(Orme and Reis 1991).
To assess our first hypotheses regarding the
Unweighted results are reported since fac-
independent effects of various social status
tors used in over-samplingwere controlled in
factors, we began by regressing our physical
all analyses and the overallpatternof findings
activity outcome on age, gender, race, educa-
were similar for both weighted and unweight-
tion, and earnings. Then, to consider how age
ed analyses (Winshipand Radbill 1994).
and gender might condition the effect of edu-
cation and earnings we added interaction terms
for gender X education, gender X earnings,
RESULTS
age X education, age X earnings, age X gender
X education, and age X gender X earnings
Social Status and VigorousPhysical Activity
interaction terms to the model (to avoid
collinearity problems, we used age, education,
Table 2 reports the results of five models
and earnings centered on the mean in the
estimatingthe impact of social status, contex-
analyses and in the construction of the interac-
tual characteristics,and personal characteris-
tion terms; Cronbach 1987). Unfortunately, the
tics on regular,vigorous exercise. Consistent
black sample was not large enough to allow for
with previous reports (e.g., U. S. Department
a consideration of higher order race interac-
of Health and Human Services 1996), the first
tions. Change in R2 was used to assess if the
model in Table 2 indicates that younger indi-
addition of the interaction terms significantly
viduals, men, non-blacks, those with more
contributed to the explanatory model (Jaccard,
education, and those with higher household
Turrisi, and Wan 1990). We then proceeded to
earnings participatein more regular physical
add blocks of contextual experiences and indi-
activity.However,Model 2 furtherreveals that
vidual-level characteristics to the social status
the effects of earningson physical activity dif-
model and evaluated the corresponding
fer by genderandthatthe educationeffects dif-
changes in the association between social sta-
fer by age and gender. Figure 2 plots the sig-
tus and physical activity.
nificant gender X earnings interaction and
Not all of the respondents were working for
TABLE 2. OLS Estimates of the Association between Social Status, Contextual Experiences,
Individual Resources and Processes, and Regular Vigorous Exercise
Model 1 Model 2 Model 3 Model 4 Model 5
-.033*** -.029*** -.025*** -.029*** -.023***
Agea
(.003)
(.002) (.004) (.004) (.004)
Gender(female = 1) -.741*** -.758*** -.708*** -.688*** -.628***
(.063) (.063) (.065) (.065) (.062)
Race/ethnicity(black = 1) -.356** -.336** -.353** -.285 -.162
(.133) (.133) (.132) (.132) (.126)
Educationa .057*** .022 .022 .022 -.006
(.013) (.018) (.018) (.018) (.017)
Household Earnings(thousands)a .003** .001 .0002 .0002 -.0003
(.001) (.001) (.001) (.001) (.001)
GenderX Agea -.007 -.009+ -.007 -.005
(.005) (.005) (.005) (.005)
GenderX Educationa .034
.065 .053* .046+
(.027) (.027) (.027) (.025)
GenderX Earningsa .003+ .004* .004* .003+
(.002) (.002) (.002) (.002)

11. 211
AND EXERCISE
SOCIALINEQUALITIES
TABLE 2. (Continued)
Model 1 Model 2 Model 3 Model 4 Model 5
Agea X Educationa .003** .003*
.003** .003*
(.001) (.001) (.001) (.001)
GenderX Agea X Educationa -.005* -.005* -.005 -.004*
(.002) (.002) (.002) (.002)
Family & Work Microsystems
ParentalStatus (Child < 18 = 1) .074 .051 .040
.077 .077 .073
.187*
Not CurrentlyMarriedb .189* .215**
(.084) (.083) (.080)
High Spouse EmotionalSupportb .185* .118 .089
(.079) (.081) (.078)
FamilyEmotional Support .120* .024 -.036
(.052) (.053) (.051)
Decision Latitudeat Workc .105** .041 .040
(.038) (.043) (.041)
HoursWorked/Week .004+ .006* .006*
(.002) (.003) (.002)
Mesosystemc& CommunityEnvironment
Work-Family
Negative SpilloverWorkto Family -.160** -.062
(.061) (.059)
Positive SpilloverWorkto Family .144** .131**
(.049) (.046)
Negative Spillover Familyto Work .046 .093
(.065)
(.068)
Positive SpilloverFamilyto Work .055 .032
(.047) (.045)
.351*** .268***
NeighborhoodSafety
(.059)
(.061)
IndividualProximalProcesses
FunctionalRestrictions -.570***
(.049)
Shortnessof Breath -.177**
(.062)
Body Mass Index -.013*
(.006)
.019*
Environmental
Mastery
(.010)
PerceivedControlover Health .071+
(.042)
.099**
Exercise Efficacy
(.036)
2.48*** 2.79***
4.52 3.40***
Constant 4.50***
.447
.045 .351
.045 .247
.253
.172
.143 .153
AdjustedR2 .138
.022*** .082***
.007*** .012***
n/a
AR2
+p <.10; * p <.05; **p < .01; ***p <.001 (two-tailed)
Source: National Surveyof MidlifeDevelopmentin the United States, 1995.
Note: Numbers in parenthesesare standarderrors.Estimatesare based on unweighteddata.
have been centeredon the mean to avoid collinearityproblems.
aVariables
bIncontrastto marriedindividualswith low emotional support.
variableswere also
only relevantto employed individuals.Missing indicatorflag variablesfor work-related
CEstimates
included in models 3-5.
activity across adulthood are the direction of
indicatesthat a higher level of household earn-
the association in early adulthoodand the rate
ings is associated with more vigorous exercise
of decline with age. In contrastto the frequent-
among women but not men.
ly reportedpositive association between more
To interpretthe significant three-way gen-
physical activity and higher education, these
der X age X education interactioneffect, we
results indicate that men with the lowest level
plotted the predicted mean levels of regular
of educationhave the highest level of regular,
exercise by age, education, and gender (see
vigorous exercise during young adulthood,
Figure 3). The most salient education differ-
while the most educated men exercise least
ences between men and women in physical

12. 212 JOURNALOF HEALTHAND SOCIALBEHAVIOR
Earningsand Gender
FIGURE2. Vigorous
Exerciseby Household
4.8 -
4.6 -
.)
4.4 -
4.2 -
0 4- Men
-I0
o
_-U-W omen
3.8 -
o
._
>
w
3.6 -
3.4 -
3.2
3 I I I
-1 S.D. Household +1 S.D. Household
Mean Household
Earnings Earnings Earnings
during the younger years. However, the age- earnings does not accumulate with age (con-
related decline in vigorous exercise is steepest traryto our cumulativeadvantagehypothesis).
among the least educatedmen and most grad- It is also noteworthythat blacks reportedsig-
ual for the best educated. This leads to age- nificantly less vigorous physical exercise than
related trajectoriesin physical activity among non-blacks,even controlling for major dimen-
men that converge in midlife (i.e., 50-55) and sions of socioeconomic status.
then proceed to provide an advantagein phys-
ical activity during later adulthood to those
with the highest education.By contrast,among ContextualCharacteristicsand Regular
women, having a high level of education is Physical Activity
associated with more vigorous exercise in
Results reported in Table 2, Model 3 also
young adulthood,but this benefit is attenuated
indicate contextualfactors from multiple con-
in later adulthood. Counter to the cumulative
texts, and multiplelevels are notablecorrelates
advantagehypothesis, the slope of age-related
decline in exercise among women is the least of regularphysical activity.In contrastto indi-
viduals in marriages characterizedby a low
steep for those with the lowest level of educa-
tion and greatest among those with the highest level of spouse emotional support,single indi-
level of education.The age X earnings,and the viduals and individuals with a high level of
gender X age X earningsinteractionterms did spouse emotional supportparticipatein more
not contribute to the model in preliminary vigorous exercise. Also consistent with our
analyses, so they are not included in Model 2. family systems-basedhypothesis, we find that
Collectively this evidence provides mixed a higher level of family emotional supportwas
support for our social status hypotheses. also independentlyassociated with more regu-
Consistent with the cumulative advantage lar exercise. Paralleling other reports using
hypothesis and our hypothesisanticipatingthat health behaviorscales as outcomes, our results
socioeconomic advantage would be most also indicatethat a higher level of decision lat-
applicableto men, we find that advancededu- itude among currently employed respondents
cation contributesto a slower decline in physi- was associated with more regular, vigorous
cal activity among men than women. However, exercise. Contrary our hypothesis,trendevi-
to
contraryto the cumulativeadvantagehypothe- dence indicate that working more hours was
sis, women with a college education have a associated with more regular exercise.
steeper decline than women with less educa- Collectively,these results supportthe ecologi-
tion. Additionally,a higher level of household cal perspective suggesting that factors from
earnings appearsto promote regular,vigorous multiple contexts of daily life influence behav-
exercise among women only (contraryto our ior and development.
gender hypothesis), and the benefit of higher Higher ordercontextualfactors, such as the

13. 213
AND EXERCISE
SOCIALINEQUALITIES
Exerciseby GenderandAge
FIGURE3. Education Vigorous
in
Men
6.00 -
5.50 -
e 5.00 -
>1fifo
C 4.50 -
o
o 4.00 -
>3.50 -
3.00 I I I I I I
I I I I I
75
25 30 35 40 60 70
45 50 55 65
Age
Women
- -< High School
5.50 -
School
--High
) 5.00 - 3 Yrs College
iquot; College Grad
4.50 -
4.00
O 4.00 -
> 3.50 -
3.00 I I I I I I I I I
I I
25 30 35 40 45 50 55 60 65 70 75
Age
Source: National Survey of Midlife Development in the United States, 1995
quality of fit between work and family and the family interface completely explained the
community environment, are also important association between family and spouse emo-
correlates of physical activity among adults. tional supportand physical activity.That is, a
Table2, Model 4 indicatesthata lower level of higher level of family and spouse emotional
negative spillover from work to family and a support is associated with more positive and
higher level of positive spillover from work to less negative spillover from work to family,
family is associated with more regular exer- which in turn is associated with more healthy
activity habits. Moreover,once the quality of
cise. Additional analyses (not shown) further
the work-family fit was controlled in the
indicated that these dimensions of the work-

14. 214 JOURNALOF HEALTHAND SOCIALBEHAVIOR
model, we find that working more, ratherthan family appearsto be indirectlyassociatedwith
fewer, hours per week was associated with lower physical activity throughits association
more regularexercise. Finally, consistent with with a lower level of environmental mastery.
our hypothesis, analyses indicatedthat respon-
dents who perceived their neighborhoods as
more safe participatedin more regular,vigor- DISCUSSION, SUMMARY,
AND CONCLUSIONS
ous exercise than individualsin less safe com-
munities. Moreover,neighborhoodsafety part-
The overall goal of this study was to use
ly explains the race/ethnicity effect; that is,
blacks are less likely than non-blacksto report ecological theory to systematically examine
the complex associations between multiple
living in a safe neighborhood,and,in turn,par-
social status characteristicsand participation
ticipate in less regularphysical activity.
in regular,vigorous exercise. We also wanted
to explore the degree to which everyday con-
ProximalResources & Processes and Regular textual factors in family, work, and communi-
ty, as well as more proximal individual
Physical Activity
resourcesandprocesses might help accountfor
Building from ecological theory,we expect- and also uniquely contributeto differences in
ed that important individual resources and exercise behaviorduringadulthood.
processes would be independentlyassociated Results from this study replicateand extend
with physical activity and might also partially previous research in several importantways.
account for the associations between social First,with regardto socioeconomic status, our
status and contextualexperiences and exercise descriptivefindings replicateothers indicating
behavior (see Table 2, Model 5). Consistent that higher levels of earnings and education
with a largebody of previousresearch,we find were both independentlyassociated with more
that physiological declines, including exercise among some populationsubgroupsof
increased body mass index, physical activity adults. Consistent with some conceptualiza-
limitations due to functional impairment,and tions of social inequality,these results suggest
shortness of breath associated with physical thathealthpractitioners need to recognizemul-
activity are associated with less participation dimensions of socioeconomic position
tiple
in regular exercise. Similarly, consistent with when formulatingand implementingstrategies
our hypotheses guided by value-expectancy to eliminate health inequalities. If socioeco-
theories of health behavior, we find that nomic status is viewed in terms of education
greater environmental mastery (a proxy for alone, interventions targeting income-related
self-efficacy), more perceived control over determinantsof physical activity will remain
health (trend), and a higher level of exercise unrecognizedor under-prioritized.
efficacy were associated with more regular Equally important however are our new
results indicating that age and gender condi-
physical activity.
The persistent race/ethnicity effect evi- tion associations between educationand regu-
denced in all previous models is completely lar exercise. This patternof results is notewor-
explained in the full biopsychosocial model of thy for several reasons. First, since the slope
exercise (i.e., Table 2, Model 5). African between educationand the rate of participation
Americansreportedhigherlevels of body mass in vigorous exercise is different depending
index, more functional limitations related to upon age and gender (Krieger et al. 1997),
activity, and greater shortness of breath than these results raise caution regardingthe unex-
whites, and this accounted for the remaining amined inclusion of educationmerely as a lin-
exercise differences between blacks and non- ear covariate in studies of physical activity.
blacks. Likewise, educational differences in These resultsalso suggest thattargetsfor phys-
physical activity among women were com- ical activity interventionsmay need to change
pletely explained by these individual-level dependingupon the age, gender,and education
resources.Womenwith a bettereducationwere of the target population. For example, inter-
in better physical condition, and had a higher ventions targetingpoorly educatedyoung men
level of each of the psychological resources may need to focus on maintaining physical
conducive to healthy exercise habits. Finally,a activity habits, while interventions targeting
higher level of negative spillover from work to poorly educated young women may need to

15. 215
SOCIAL
INEQUALITIES EXERCISE
AND
thatprovidethe flexibility,opportunity, the
focus on initiating exercise habits. Finally, and
expectation of regular physical activity (e.g.,
these results again remindscholarsthat survey
questions about physical activity may tap dif- golf outings with clients, on-site fitness facili-
ferent meanings for different groups of indi- ties, company sponsoredgym memberships).
viduals. That is, it is possible that the counter- The contextual effects found in this study
intuitiveeducation-exerciseassociationamong extend previous health behavior research.
young men is capturing occupation-based First,in contrastto several studies using health
behavior indices, this study found that differ-
physical activity that may yield less notable
health gains than regular aerobic-forms of ent aspects of work and family were specifi-
exercise, which may have more health bene- cally associatedwith physical activity.Perhaps
fits. even more importantly,however, this study
We also think that the explanatorypower of demonstrated how overlookedaspects of work
the social status model is worthy of attention. and family may influence different health-
Although our final model only accounted for relatedbehaviors.For example, the work-fam-
25 percentof the variancein vigorous exercise, ily literaturehas been dominated by studies
social status factors provided most of the exploring the effects of work-family conflict
explanatorypower (i.e., 14 percent). Consis- (Barnett 1998), and this research frequently
tent with an ecological perspective, this finds that work-family conflict is associated
evidence suggests that models that integrate with behaviors and conditions (e.g., alcohol
theory and concepts from across disciplines abuse and depression; Frone, Russell and
Cooper 1997) that undermineregularexercise
(e.g., social status and behavioral intention)
will generate more comprehensive explana- (King et al. 1992). However, the evidence
tions for behaviorand may facilitatemore suc- reportedin this study indicate that more posi-
cessful interventionsto change behavior. tive spilloverfromworkto family (a dimension
not typically considered) has the most robust
It is also interestingto note that contextual
association with greater frequency of regular
experiences and physiological resources
explained education-level differences in exer- physical activity.
The pattern of findings surrounding
cise among youngerwomen. These resultspar-
race/ethnicityin these models is also an impor-
tially supportthe hypothesis that social status
The
conditions the resources and experiences that tant contributionto the exercise literature.
fact that blacks were consistently found to
facilitate participationin positive health-relat-
exercise less than non-blacks, independentof
ed behaviors (e.g., Ross and Wu 1995). This
education and earnings, suggests that race is
everyday interactional perspective of the
not an adequateproxy for socioeconomic sta-
socioeconomic statusand healthbehaviorrela-
tus, and thatotherfactorspertainingto race are
tion is markedly different from assumptions
contributingto differences in physical activity
regarding the absence of important health
patterns. Consistent with a recent biopsy-
related benefits of exercise (e.g., the Health
chosocial model of racism as a stressor(Clark
Behavior Model) that may accompanyhigher
et al. 1999), our results suggest that contextual
levels of education. Although the pattern of
factors from the social environment such as
results for education among men supportsthe
neighborhoodsafety may directlyand indirect-
cumulative advantage hypothesis and theory
ly undermineexercise habits. Inadequatestreet
suggesting that the health effects of socioeco-
lighting at a modest level or witnessing violent
nomic status are gendered (Marmot et al.
crime in your neighborhood at an extreme
1987; McDonough et al. 1999), none of the
level are direct physical barriers to outdoor
contextual experiences or individual-level
regular,vigorous activity, while the threat of
resources could explain men's educationaltra-
victimization (e.g., Ross 1993) and compro-
jectories across adulthood.Perhapsthe persis-
mised feelings of masteryor controlare impor-
tent educationeffect among men can be attrib-
tantpsychological barriers.These physical and
uted to the confounded nature of physical
psychological barriersto physical activity may
activity in men's occupations(see Eakin 1997).
set into motion reductionsin physical activity
That is, men who have a low education may
habitswhich become furtherreinforcedby cor-
find themselves in working-class jobs that
responding reductions in physiological func-
become less physically demanding as they
tioning and ultimately undermine health.
spend more time on the job, while men with
Althoughthese linkages are supportedtheoret-
more education may work their way into jobs

16. 216 JOURNALOF HEALTHAND SOCIALBEHAVIOR
ically (Clark et al. 1999), we could not evalu- study may be due to common methodvariance
ate these causal pathways with the current or endogeneity.Additionally,a more stringent
cross-sectional data. operationalizationof ecological theory would
We also feel the persistence of the robust require measures from additional levels and
age effect on exercise should not go unno- sources (e.g., self-report of decision latitude
ticed. Overall, in these data about 32 percent supplementedby a rating of decision latitude
of respondents aged 25-35 indicated they for a respondent's occupational category).
participated in vigorous exercise several Moreover,a comprehensiveecological model,
times a week or more, while 28 percent of informed by other middle range theories,
those 45-54, and 18 percent of those aged might furtherinclude a myriadof other possi-
65-74 did. After adjusting for so many con- ble interactions.
textual and individual factors, including Nonetheless, our results supportthe value of
declines in physiological capacity, we might
employing an ecological perspective when
have expected this effect to diminish more
attemptingto understandhealth behavior.We
substantially. find that, indeed, different aspects of social
In an attemptto furtherunpack the persis- status were independently associated with
tent age effect in these analyses,we rana series
exercise, and that the influence of some status
of post-hoc analyses including two- and three- factors variedby gender and age. We also find
way interactions terms between age, gender, that contextualexperiencesin family and work
and the contextual correlates. Only one of all
life, the quality of the work-family interface,
the possible interactionterms emerged signifi- and the community environment were all
cant; paralleling other reports (Potts et al.
directly or indirectly associated with exercise
1992; Rakowski et al. 1987; Zimmermanand habits and added significantly to the explana-
Connor 1989), the two-way interaction
tory power of our model. In short, our results
between age and family emotional supportwas
supporta biopsychosocialperspectiveof phys-
significant (p < .05), suggesting that family ical activity duringadulthood,and they suggest
supportwas associatedwith more regularexer- that future health behavior research needs to
cise among older adultsthanyoungeradults.In
consider determinantsfrom multiple contexts
all the post-hoc models we examined,the neg-
and multiple levels of the humanecology.
ative associationbetween age and exercise per-
sisted. What other experiences underminepar-
ticipation in regularexercise as adults traverse
REFERENCES
the life course?
We acknowledge the limits of these cross-
Albert, Steven M. 1995. quot;New Perspectiveson the
sectional data when interpretingage-effects in
Elderly: Epidemiology and Public Health.quot;
our analyses. That is we cannot differentiate CurrentIssues in Public Health 1:77-81.
true quot;agequot;(i.e., developmental)from quot;cohortquot; American College of Sports Medicine. 1990. quot;The
or quot;periodquot; effects. For example, the oldest RecommendedQuantityand Qualityof Exercise
group in this sample would have been at the for and
Developing Maintaining
height of midlife when the Surgeon General Cardiorespiratory and Muscular Fitness in
announcedthe first Healthy People objectives HealthyAdults.quot;Medicine and Science in Sports
in 1979. Since original messages about exer- and Exercise 22:265-74.
cise during this period targeted reducing car- Backett, Kathryn. 1992. quot;The Construction of
HealthKnowledge MiddleClass Families.quot;
in
diovasculardisease (particularlyamong men),
Health Education Research: Theory and
it might be that highly educated young and
Practice 7:497-507.
middle-aged men from these cohorts respond-
Backett, Kathryn C. and Charlie Davison. 1995.
ed to such social marketingand this is reflect-
quot;Lifecourse and Lifestyle: The Social and
ed in contemporaryage-relatedphysical activ-
CulturalLocation of Health Behaviours.quot;Social
ity patterns. Not all birth cohorts might be Science and Medicine 40:629-38.
expected to reportthe same levels of physical Barett, Rosalind C. 1998. quot;Towarda Review and
activity at young and old ages. Reconceptualization of the Work/Family
Considerationof these results must also be Literature.quot; Genetic, Social & General
temperedby additionallimitationsof the study. Psychology Monographs124:125-82.
For example, the data were all self-reported; . 1999. quot;A New Work-Life Model for the
consequently,some of the associations in this Annals of the American
Twenty-FirstCentury.quot;

17. 217
AND
SOCIAL
INEQUALITIES EXERCISE
Academy of Political and Social Science LaPorte, and Nancy E. Warhaftig. 1991.
562:143-58. quot;Physical Activity by Socioeconomic Status in
Becker, MarshallH. 1986. quot;The Tyrannyof Health Two PopulationBased Cohorts.quot;Medicine and
Promotion.quot; Public Health Reviews 14:15-25. Science in Sports and Exercise 23:343-52.
Berkman,Lisa F. and Lester Breslow. 1983. Health Centers for Disease Control. 1997. quot;Stateand Sex-
and Waysof Living: TheAlameda CountyStudy. Specific Prevalenceof Selected Characteristics:
New York:Oxford UniversityPress. Behavior and Risk Factor Surveillance System,
Bielby, Denise D. 1992. quot;Commitment Workand 1994 and 1995.quot;Morbidityand MortalityWeekly
to
Family.quot; Annual Review of Sociology Report46:1-32.
18:281-302. . 1999. quot;Neighborhood Safety and the
Blair, Steven N. 1984. quot;How to Assess Exercise Prevalence of Physical Inactivity: Selected
Habits and Physical Fitness.quot; Pp. 424-47 in States, 1996.quot; Morbidity and Mortality Weekly
Behavioral Health: A Handbook of Health Report 48:143-46.
Enhancementand Disease Prevention,edited by Cheadle, Allen, Bruce M. Psaty, Paula Diehr,
Joseph D. Matarazzo,Sharlene M. Weiss, J. A. Thomas Koepsell, Edward Wagner, Thomas
Wickezer, and Susan Curry. 1999. quot;An
Herd, Neal E. Miller, and Stephen M. Weiss.
New York:JohnWiley & Sons. Empirical Exploration of a Conceptual Model
for Community Based Health Promotion.quot;
. 1988. quot;Exercise Within a Healthy
International Quarterly of CommunityHealth
Lifestyle.quot;Pp. 75-89 in ExerciseAdherence:Its
Impact on Public Health, edited by Rod K. Education 13:329-63.
Chilcoat, Howard D. and Naomi Breslau. 1996.
Dishman. Champaign, IL: Human Kinetics
quot;AlcoholDisordersin YoungAdulthood:Effects
Books.
of Transitions into Adult Roles.quot; Journal of
Blaxter, Mildred. 1990. Health and Lifestyles.New
Health and Social Behavior 37:339-49.
York:Routledge.
Clark, Rodney, Norman B. Anderson, Vernessa R.
Brazier,J. E., R. Harper, M. Jones,A. O'Cathain,
N.
Clark,andDavid R. Williams. 1999. quot;RacismAs
K. J. Thomas, T. Usherwood, and L. Westlake.
a Stressor for African Americans.quot; American
1992. quot;Validating the SF-36 Health Survey
Psychologist 54:805-16.
Questionnaire: New Outcome Measure for
Cronbach, Lee J. 1987. quot;Statistical Tests for
Primary Care.quot; British Medical Journal
Moderator Variables: Flaws in Analyses
305:160-64.
Recently Proposed.quot; Psychological Bulletin
Breslow, Lester and Norman Breslow. 1993.
quot;Health Practices and Disability: Some 102:414-17.
Dishman, Rod K. 1991. quot;Increasing and
Evidence From Alameda County.quot; Preventive
Maintaining Exercise and Physical Activity.quot;
Medicine 22:86-95.
Behavior Therapy22:345-78.
Bromet, Evelyn J., Mary A. Dew, and David K.
Doherty, William J. and Jill E. Harkaway. 1990.
Parkinson. 1990. quot;Spillover between Work and
quot;Obesity and Family Systems: A Family FIRO
Family:A Studyof Blue-CollarWorkingWives.quot;
Approach to Assessment and Treatment
and Family,
Pp. 133-52 in Stress Between Work
Planning.quot; Journal of Marital and Family
edited by John Eckenrodeand Susan Gore. New
Therapy16:287-98.
York:Plenum Press.
Doherty, William J. and Hamilton I. McCubbin.
Urie. 1979. TheEcology of Human
Bronfenbrenner,
1985. quot;Familiesand Health Care:An Emerging
Development: Experiments by Nature and
Arena of Theory, Research, and Clinical
Design. Cambridge, MA: Harvard University
FamilyRelations 34:5-11.
Intervention.quot;
Press.
Dunn, Andrea L., Bess H. Marcus, James B.
. 1995. quot;Development Ecology Through
Kampert, Melissa E. Garcia, Harold W Kohl,
Space and Time: A Future Perspective.quot; Pp.
and Steven N. Blair. 1999. quot;Comparison of
619-47 in Examining Lives in Context:
Lifestyle and Structured Interventions to
Perspectives on the Ecology of Human
IncreasePhysicalActivity and Cardiorespiratory
Development, edit by Phyllis Moen, Glen H.
Fitness: A Randomized Trial.quot;Journal of the
Elder, and Kurt Luscher. Washington, DC:
AmericanMedicalAssociation 281:327-34.
AmericanPsychological Association.
Dzewaltowski, David A. 1994. quot;Physical Activity
Bronfenbrenner,Urie and Stephen J. Ceci. 1994.
Determinants:A Social Cognitive Approach.quot;
in
quot;Nature-Nurture Reconceptualized
Medicine and Science in Sports and Exercise
Developmental Perspective: A Bioecological
26:1395-99.
Model.quot;Psychological Review 101:568-86.
Determinants
Eakin, Joan M. 1997. quot;Work-Related
Urie and PamelaA. Morris. 1998.
Bronfenbrenner,
of Health Behavior.quot; 337-58 in Handbookof
quot;The Ecology of DevelopmentalProcesses.quot;Pp. Pp.
Health Behavior Research: Personal and Social
993-1028 in Handbookof ChildPsychology, 5th
Determinants, vol. 1, edited by David S.
ed. vol. 1, edited by W. Damon. New York:John
Gochman.New York:Plenum Press.
Wiley & Sons.
Eyler,Amy A., ElizabethBaker,LaChennaCromer,
Cauley, Jane A., Sharyne M. Donfield, Ronald E.