| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
TOPIC 4. ECONOMIC STATUS AND HEALTH INEQUALITIES |
1 Pennsylvania State University, University Park.
2 Institute for Social Research, University of Michigan, Ann Arbor.
Address correspondence to Linda A. Wray, PhD, Department of Biobehavioral Health, Pennsylvania State University, 315 Health and Human Development East, University Park, PA 16802. E-mail: law30{at}psu.edu
Abstract
Objectives. We focus on a hypothesized mechanism that may underlie the well-documented link between social status and health–behavioral health risks.
Methods. We use longitudinal data from representative samples of 6,106 middle-aged and 3,636 older adults from the Health and Retirement Study to examine the relationships between social status—including early life social status (e.g., parental schooling), ascribed social status (e.g., sex, race–ethnicity), and achieved social status (e.g., schooling, economic resources)—and behavioral health risks (e.g., weight, smoking, drinking, physical activity) to (1) assess how early life and ascribed social statuses are linked to behavioral health risks, (2) investigate the role of achieved factors in behavioral health risks, (3) test whether achieved status explains the contributions of early life and ascribed status, and (4) examine whether the social status and health risk relationships differ at midlife and older age.
Results. We find that early life, achieved, and ascribed social statuses strongly predict behavioral health risks, although the effects are stronger in midlife than they are in older age.
Discussion. Ascribed social statuses (and interactions of sex and race–ethnicity), which are important predictors of behavioral health risks even net of early life and achieved social status, should be explored in future research.
SOCIAL status—which helps shape the life course pathways followed by individuals and, in turn, their exposures to experiences that promote or prevent healthy development—can be conceptualized in terms of both achieved and ascribed positions. Achieved positions are those gained through access to opportunities and exercise of individual volition, whereas ascribed positions are based on biological or group attributes acquired at birth. Social scientists have long argued that achievements, as measured primarily by socioeconomic status (e.g., education, occupational status, and economic well-being), play an important role in understanding inequalities in health (e.g., Adler et al., 1994
; Antonovsky, 1967; Ettner, 1996; Feinstein, 1993; House, Lepkowski, Kinney, Mero, Kessler, & Herzog, 1994; Kitagawa & Hauser, 1973; Ross & Wu, 1995, 1996; Syme & Berkman, 1976). Decades of studies have documented that higher achieved statuses lead to increased access to health care, fewer diagnosed diseases and impairments, lesser severity of diseases and impairments, reduced doctor and hospital visits, and lower health care costs. Studies have revealed that health disparities also vary by ascriptive categories such as sex and race–ethnicity, independent of achieved statuses (Adler et al., 1994; Crimmins, Hayward, & Seeman, 2004; Ferraro & Farmer, 1996; Hayward & Heron, 1999; Hayward et al., 2000; Hummer, Rogers, Nam, & LeClere, 1999; Whitfield & Seeman 1997; Williams & Collins, 1995; Williams, Jackson, & Anderson, 1997; Winkelby, Kraemer, Ahn, & Varady, 1998).
There are several lines of research undertaken to systematically document what factors may explain the effects of achieved and ascribed social status on health, how they interrelate, and whether patterns differ by disease or age group. Some efforts aim to document a spurious relationship between social status and health. For example, early life factors and later experiences may be the explanation for the role of sex and race disparities in health (Blackwell, Hayward, & Crimmins, 2001; Hayward & Gorman, 2004; Wray & Blaum, 2001). Still other researchers proposed to understand the links as a social and cultural phenomenon, examining mediational factors such as behavioral health risks that may explain the link (Haan, Kaplan, & Camacho, 1987; Himes, 2000; House et al., 1994; Lantz, House, Lepkowsky, & Williams, 1998; Winkelby, Fortmann, & Barrett, 1990). These and other studies (e.g., Jenkins, Fultz, Fonda, & Wray, 2004; Wray, Blaum, Ofstedal, & Herzog, 2004) have shown considerable variation by age, sex, race–ethnicity, and achieved status in behavioral health risks such as obesity, physical activity, heavy drinking, and smoking. However, recent studies testing whether health behaviors themselves mediate the achieved status–health link have found only modest effects (Crimmins et al., 2003; House et al., 1994; Lantz et al., 2001; Lynch, Kaplan, & Salonen, 1997). Because behavioral health risks are largely modifiable, additional investigations of their influence on the social status–health link over time—and whether the patterns differ by ascribed status and across age groups—are nonetheless warranted, given their considerable research, policy, and intervention relevance.
RESEARCH OBJECTIVES
Our goal is to answer four questions on the relationships between multiple measures of early life, ascribed, and achieved social statuses and behavioral health risks in middle-aged and older U.S. adults: (1) How are early life and ascribed social statuses linked to behavioral health risks in middle-aged and older adults? (2) How do achieved social statuses affect variation in risky health behaviors? (3) Do achieved statuses explain the effects of early life and ascribed social status? (4) Do the relationships between early life, achieved, and ascribed social statuses and health risks differ at midlife and older age?
DATA AND MEASURES
Samples
The original Health and Retirement Study (HRS) and Study of Assets and Health Dynamics (AHEAD) are large nationally representative surveys of middle-aged and older adults that are well suited to answering the study's research questions. The surveys include information on major diseases and impairments experienced by Americans in these age groups and multiple measures of social status as well as other factors associated with both health and social status; and the panel designs permit both cross-sectional and longitudinal analyses. The original HRS began to collect data on 9,824 adults aged 51–61 years in 1992 (and their spouses or partners of any age), and the AHEAD began to collect parallel data on 7,773 adults aged 70 years and older (and their spouses/partners) in 1993, with oversampling of Black and Latino Americans and reinterviews every 2 years after baseline with good response rates (82% in 1992 and 89% in 1994, HRS; 80% in 1993 and 93% in 1995, AHEAD). The two surveys were integrated in 1998 (retaining the HRS name) when respondents in both of the original age panels were reinterviewed. Further details are available elsewhere (Juster & Suzman 1995; Soldo, Hurd, Rogers, & Wallace, 1997).
This study employed data from two HRS waves (1992, 1994) and two AHEAD waves (1993, 1995) to test the relative contribution of various social status measures to behavioral health risks in midlife and older age (and, ultimately, to 1998 health outcomes in a larger study based on these data plus data from the combined 1998 wave). Our samples included 6,106 middle-aged and 3,636 older respondents with complete data at baseline and reinterviews. (Because this study is part of a larger study examining the effects of social status at baseline [1992 for HRS, 1993 for AHEAD] and subsequent behavioral health risks [in 1994–1995 for HRS and AHEAD, respectively] on comorbidity in 1998, the samples included 6,106 middle-aged adults and 3,636 older adults drawn from cases with complete data at baseline and follow-up interviews in 1994 and 1995 as well as 1998. Given that attrition (due to mortality, refusals, or other reasons) results in reinterviews with respondents who are, on average, healthier and more economically secure than are nonrespondents, our results may underestimate the effects of social status on health risks). We took advantage of the longitudinal designs to establish a causal ordering of the model variables, obtaining data on social status at baseline in each study and risky health behaviors from the 1994–1995 reinterviews for this article.
Measures
To compare age groups, we restricted the analytic variables to those that were identical (or nearly so in one case, as noted below) in both surveys. Our dependent variables were polychotomous measures of behavioral health risks representing riskier categories of body weight, drinking, smoking, and physical activity versus the least risky category, according to current classifications of health risks for adults. Although the risk thresholds may arguably differ somewhat for middle-aged and older adults, we used equivalent criteria for the behavioral health risks to compare patterns between the age groups. In particular, risky weight was measured as underweight (body mass index < 19) or obese (body mass index > 30) versus normal or overweight. Drinking status was measured as being a nondrinker (0 alcoholic drink/day) or a heavy drinker (3 or more drinks/day) versus a moderate drinker (1–2 drinks/day). Smoking behavior was assessed as being a current smoker or past smoker versus a never-smoker. Finally, low physical activity was measured as not participating in vigorous exercise at least three times a week versus being a vigorous exerciser.
Independent variables included five early life, achieved, and ascribed social statuses shown in previous studies to be associated with increased risk for disease. First, we included one measure of early life social status attributable to one's family of origin that hypothetically influences later achieved social status: parental education, measured as the average of father's and mother's years of schooling. In the case of the middle-aged adults, parental schooling is a continuous measure of the average years of schooling (range 0–17) reported for the respondent's mother and father. Older adults were asked whether neither, one, or both of their parents had at least an eighth grade education (range 0–2). (Because the measures for parental schooling differed across the age groups, we tested whether continuous versus categorical measures in the HRS would result in different estimates of the effects of parental schooling on behavioral health risks. First, we collapsed the continuous measures of schooling for respondent's mother and father into categories equivalent to those in the AHEAD. Second, we ran our models using both continuous and categorical measures of parental schooling and found no significant differences in the effects of either measure. Because we found no difference, we chose to use the continuous measure of parental schooling for the HRS sample to allow for finer variation in our estimates.)
We can imagine that a more advantaged upbringing fosters early life modeling of protective health behaviors engaged in by one's parents. Second, we included two measures of achieved social status—education (respondent schooling attained in years, range 0–17) and current economic resources (standardized sum of baseline income and assets)—to test if different achieved statuses vary in their effects on health risks. For example, one's own education may contribute to risks for and complications of disease by increasing knowledge about health behavior–disease links (Wray, Herzog, Willis, & Wallace, 1998), and economic resources represent accumulated financial resources that allow one to "buy" better health in general and behavior modification programs, medications, or other treatments in particular.
Third, ascribed social status was measured as sex (0 = male, 1 = female), Black (0 = White non-Latino, 1 = Black/African American), and Latino (0 = White non-Latino, 1 = Hispanic/Latino American). We excluded cases of respondents who self-identified as Asian American or American Indian because they represented too few cases to derive reliable estimates and because disease prevalence rates differ markedly for these groups compared with other major race–ethnicity groups.
Finally, we included age and marital status in our models. We examined the influence of age in two ways: (1) within-group differences in chronological age (range 51–61 years, HRS; 70+ years, AHEAD); and (2) age group differences that may reflect survivorship. Being married/partnered versus other marital states was included as a measure of "social control" shown to be related to risky health behaviors (Franks, Pienta, & Wray, 2002; Umberson, 1987, 1992).
RESULTS
Table 1 describes the HRS and AHEAD samples across these demographic, social status, and risky health behavior characteristics (overall and by sex within each age group), providing evidence of patterns similar to those based on data from other large epidemiological studies of health. For example, although middle-aged and older adults are relatively healthy on average, older adults experience somewhat poorer health than adults at midlife. The older adults are less likely to be married and report fewer years of schooling and lower levels of economic resources compared with their middle-aged peers. There is even wider variation by sex in each age group. Women are poorer and less healthy and report fewer years of schooling compared with men; older women are particularly disadvantaged on the economic measures relative to all other groups. Men and women also differ in risky health behaviors: Higher proportions of women are underweight or obese or participate in low physical activity compared with men, whereas more men than women are smokers (particularly past smokers) and heavy drinkers (particularly in middle age).
|
|
Table 3 presents results of similar analyses for being a current or past smoker versus never-smoker. Notable are strong effects of sex and marital status: Women and married adults in both age groups are far less likely to be current or past smokers compared with men and unmarried adults. Overall, higher achieved status decreases smoking behavior in both midlife and older age; however, higher levels of parental education increases the odds of being a past smoker in midlife and current smoker in older age. As in the earlier analyses, early life and achieved status account for little of the sex or marital status effects on smoking.
|
|
|
DISCUSSION AND CONCLUSIONS
Our results add to the literature on antecedents of behavioral health risks by providing further support for the hypothesis that social status produces strong and consistent effects on those risks. Not only do early life and achieved status characteristics generate substantial effects on behavioral health risks overall—and some more than others—but so does ascribed social status, even net of early life and achieved status. Although the effects of some social status factors on health risks—for example, Black status—can be explained partly by socioeconomic disadvantage, in general the ascribed status effects remain independent of early life and achieved status; and all social status factors have strong reduced-form effects on health risks in both groups, although the effects of all variables are weaker (by about half) in older age. The lesser effects for older adults may be attributable to the posited biological leveling of the population; that is, that effects of social status become less important in older age (e.g., see Beckett, 2000).
Although the study provides further evidence for effects of social status on behavioral health risks (both broadly defined), the effects of different measures of social status on the individual health risks are worth noting as well. For example, none of the social status factors except sex predicted underweight, but all social status factors predicted obesity. Higher levels of parental and respondent schooling decreased the odds of being obese in both midlife and older adulthood, and greater economic resources lowered the odds of being obese for older adults, over and above the schooling measures. Finally, being a middle-aged Black woman or an older woman in general strongly increased those odds. Overall, smoking behavior was unaffected by race–ethnicity but was strongly affected by sex. Being female strongly predicted both past and current smoking behavior, net of early life and achieved social status. Early life social status exerted little influence on past smoking behavior in older adults but played a role in increasing the odds of being a past smoker in midlife. In contrast, drinking behavior was heavily influenced by both sex and race–ethnicity. Women overall in both age groups and Latino women in middle age were routinely nondrinkers or light drinkers. The very few cases of heavy drinking in older Latinos in general resulted in odds of nearly zero. In all cases, both early life and achieved social status reduced the odds of being a nondrinker but did little to explain the ascribed status effects. Achieved status decreased the odds of being a heavy drinker only in midlife, scarcely changing the effect of being female. Sex but not race–ethnicity was a strong predictor of lower levels of physical activity in both age groups. Both higher levels of parental education and economic resources decreased the odds of engaging in low physical activity.
Understanding the mechanisms put in place by differences in social status that underlie differences in risky health behaviors—what operate independently, what are explained by others, and how they might interrelate—is clearly important to developing policies and interventions aimed at redressing the negative effects of a disadvantaged upbringing in midlife or older age. The idea that "poor people behave poorly" is clearly oversimplified. Although we find that there is a link between social inequalities and disparities in health behavior, we suspect that these differences need to be understood from a social environmental perspective. Such a perspective would emphasize the structural and social constraints on behavior, differences in opportunities for healthful behavior, and the fundamental social conditions that shape choices of social pathways that ultimately promote and/or prevent health risks (Link & Phelan, 1995).
Several characteristics of the HRS data set strengthened our ability to examine the social status–health risk links. First, because of strong measurement design, we were able to construct multiple measures of early life, achieved, and ascribed social status. Similarly, we were able to include multiple measures of behavioral health risks (as well as multiple measures of behavior-related health status, not presented in the current study). Second, the longitudinal data allowed us to specify a causal model to test the effects of social status on subsequent behavioral health risks.
We also acknowledge our study's limitations. First, our behavioral health risk measures were self-reported and relatively superficial rather than clinically derived and more in depth. Second, despite including several behavioral health risk measures, we did not examine the effects of social status on change in those risks. One can imagine that higher achieved status affects health by fostering changes in health risks—for example, weight loss or change in diet—in response to recent diagnoses or other motivating factors that have salutary effects on subsequent health status (e.g., see Wray et al., 1998, 2004). Third, our results might have differed had we made a different decision about using equivalent measures of health risk thresholds for middle-aged and older adults. Finally, we were not able to test for coping skills, self-efficacy, stress-management resources, discrimination, allostatic load, or neighborhood effects that may be associated with behavioral health risks in particular as well as wider social status–health links in general (McEwen & Seeman, 1999; Mirowsky & Ross, 2003; Robert & Lee, 2003; Whitfield, Weidner, Clark, & Anderson, 2002).
The age differences we observe in the link between social inequalities and health behaviors (i.e., our finding of less predictability of health risk behaviors in the AHEAD versus the HRS sample) remains somewhat of a puzzle. Although evidence exists for increasing inequalities in both income and health over the life span (Dannefer, 2003), the level of the association is generally found to decline in older age (e.g., House et al., 1994; Marmot & Shipley, 1996; Robert & House, 1996). Our findings parallel these results. However, as we argue elsewhere (see Alwin & Wray, this issue), rather than our findings refuting the hypothesis of the effects of social status on health, the leveling of its effect in older age is probably just the opposite. Research on aging, social status, and health risk behaviors has yet to address this question. To the extent that these and other problems can be studied in future research, we may illuminate the pathways by which social inequalities are translated into health disparities.
Acknowledgments
This study was supported by Grant No. AG15437 from the National Institute on Aging (Duane F. Alwin, principal investigator, and Linda A. Wray, co-principal investigator).
The authors thank Timothy Manning and Pauline Mitchell for their programming and graphics assistance.
This article is dedicated to A. Regula Herzog and Laurie M. Staples, extraordinary colleagues and friends.
References
This article has been cited by other articles:
 |
|
 |
|
  T. R. Bertoldo Benedetti, L. H. Goncalves, E. Petroski, S. M. Nassar, A. Schwingel, and W. Chodzko-Zajko Aging in Brazil: Physical Activity, Socioeconomic Conditions, and Diseases Among Older Adults in Southern Brazil Journal of Applied Gerontology, November 1, 2008; 27(5): 631 - 640. [Abstract] [PDF]
|
|
|
|
 |
|
 L. A. Wray, D. F. Alwin, R. J. McCammon, T. Manning, and L. E. Best Social status, risky health behaviors, and diabetes in middle-aged and older adults. J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2006; 61(6): S290 - S298. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
