The Stigma of Obesity

Home , Class discrimination, Height discrimination

Social Psychology Quarterly 74(1) 76–97 The Stigma of Obesity: Ó American Sociological Association 2011 DOI: 10.1177/0190272511398197 Does Perceived Weight http://spq.sagepub.com Discrimination Affect Identity and Physical Health?

Markus H. Schafer1 and Kenneth F. Ferraro1

Abstract Obesity is widely recognized as a health risk, but it also represents a disadvantaged social position. Viewing body weight within the framework of stigma and its effects on life chances, we examine how perceived weight-based discrimination influences identity and physical health. Using national survey data with a 10-year longitudinal follow-up, we consider whether perceptions of weight discrimination shape weight perceptions, whether perceived weight discrimination exacerbates the health risks of obesity, and whether weight perceptions are the mechanism explaining why perceived weight discrimination is damaging to health. Perceived weight discrimination is found to be harmful, increasing the health risks of obesity associated with functional disability and, to a lesser degree, self-rated health. Findings also reveal that weight-based stigma shapes weight perceptions, which mediate the relationship between perceived discrimination and health.

Keywords obesity, stigma, discrimination, health

The sense that one has been treated weight discrimination). Though less unfairly at work or in public places frequently studied, social reactions to can have negative consequences for body weight may be linked to opportu- sentiment and health. When discrimi- nity structures and personal well- nation is perceived to be related to being, but the mechanisms for how race or ethnicity (an ascribed status), this occurs are a matter of ongoing it is often viewed as an overt form of debate (Muennig 2008; Puhl and racism, initiating a stress process Brownell 2001). that may compromise physical and mental health (Gee 2002; Williams, Neighbors, and Jackson 2003; Taylor 1Purdue University and Turner 2002). Other forms of perceived discrimination, however, may Corresponding Author: Markus Schafer, Center on Aging and the Life be linked to attributes or conditions Course, Purdue University, Young Hall, 155 S. that are developed over time, such as Grant St, West Lafayette, IN 47907-2114 the case with excess body weight (i.e., Email: [email protected]

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 77

There is some evidence from studies affects health, this study examines of perceived ethnic discrimination that body weight, self-perceptions of weight positive ethnic identity plays a protective status, perceived weight discrimination, role on mental health (Mossakowski and changes in health observed over 10 2003; Sellers et al. 2003), but it is years. We posit that the social processes unlikely that a parallel effect would be involved in identity formation and revi- widely observed from weight-based dis- sion are critical to assessing how physi- crimination. Despite social movements ological factors such as excess body for self-acceptance among overweight weight influence health. Missing from people, relatively few people favorably much of the previous literature is a con- identify themselves as overweight or sideration of how interpretive processes obese. Rather, many feel that being fat associated with body weight shape is a stigmatizing experience, one that health outcomes. We therefore draw limits social and economic opportunities from studies of stigma to examine the and operates as the ‘‘last acceptable antecedents and health consequences basis of discrimination’’ (Puhl and of identity as an overweight person.2 Brownell 2001:788). As a basis of One expects that excess body weight stigma, heavy body weight constrains is related to identification as an over- or harms occupational chances, delivery weight person, but we ask whether per- of health services, educational attain- ceived weight discrimination heightens ment, family relations, self-concept, identity as an overweight person. The and various indicators of well-being analysis aims not only to document (Carr and Friedman 2005; Carr how discrimination may get ‘‘under and Friedman 2006; Crosnoe 2007; the skin’’ in a medical sense (McEwen Crosnoe, Frank, and Mueller 2008; 1998), but also to examine whether Puhl and Brownell 2001). perceived discrimination shapes the Increasingly, even medical resea- way people evaluate themselves in rchers are acknowledging that excess relation to a stigmatized condition. body weight is harmful for health not We anticipate that people do not inter- solely on a physiological basis, but in pret their body weight status on the part because of the stress associated basis of mere physiology, but rather with enduring an unfavorable social through interaction with others. trait (Muennig 2008).1 Implicit in Thus, we approach the analysis from this proposition are core sociological a sociological understanding of stigma. themes, including the internalization of stigma and interpretive self- identity processes borne out of social interactions. To more clearly understand how 2Throughout the paper, we use the term ‘‘over- perceived weight-based discrimination weight’’ when referring to identities, but speak of ‘‘obesity’’ only in terms of its official classification, which is a body mass index of 30 or more. A main 1The reasons for perceived discrimination’s reason that we do not mention ‘‘obese’’ identities deleterious effects on health are many, including owes to the wording of the survey questions stress of repeated contact with antagonistic that refer to ‘‘very overweight’’ and ‘‘somewhat others, rejection or avoidance in social settings, overweight.’’ Moreover, in colloquial discourse, negative self perceptions, and differential alloca- people more commonly refer to being somewhat tion of resources via social segregation or very ‘‘overweight,’’ rather than making (Campbell and Troyer 2007; Carr and Friedman refined, diagnostic categorizations of themselves 2005; Krieger 1999; Muennig et al. 2008). and others.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 78 Social Psychology Quarterly 74(1)

STIGMA AND BODY WEIGHT Most studies on stigmatization in sociology focus on ‘‘unusual condi- In discussing distinctive features of the tions,’’ such as severe mental illness self, Goffman (1963) developed a nuanced (e.g., Schulze and Angermeyer 2003) depiction of stigma and articulated how or HIV/AIDS (e.g., Parker and it plays out in everyday social life. Aggleton 2003). This interest in the Stigmatized traits, or ‘‘deeply discredit- ‘‘unusual’’ has been noted as an ing’’ characteristics, make people appear arbitrary boundary mechanism that dangerous or unacceptable in the eyes of sets apart stigma scholarship from its others, reduce their life chances, and iso- close intellectual cousins—prejudice late them (Goffman 1963:3). In his clas- and discrimination—which focus on sic statement on the subject, Goffman usual traits such as race, gender, or (1963:3) argued that people are rejected religion (Stuber, Meyer, and Link and classified as undesirable on three 2008). Indeed, some scholars assert accounts: (1) ‘‘tribal stigmata,’’ (2) that prejudice and stigma stem from ‘‘abominations of the body,’’ and (3) a singular theme, leading Phelan ‘‘blemishes of individual character.’’ Not and colleagues (2008) to suggest the only do outsiders look down upon the latter term be used as the unifying stigmatized individual, but the victim concept from which flow attitudinal him- or herself absorbs the discomfort and behavioral responses (i.e., preju- and unease from social interaction into dice and discrimination, respectively). his or her own self-concept. Therefore, the experience of perceived Soon after Goffman’s book, other weight discrimination will be treated sociologists began to show that body in this paper as an indication of weight—obesity in particular—is an stigma. attribute related to embarrassment and ultimately to life chances (Cahnman 1968; Maddox, Back, and OBESITY’S EFFECT ON HEALTH Liederman 1968). Several classic studies showed that not only was corpulence Though we are mainly interested in seen as a physical blight, but also that how issues related to stigma are associ- body weight is thought to be under peo- ated with health declines, it is useful ple’s control; thus, obese people were fre- from the outset to anticipate that the quently considered lazy, self-indul- physiology of heavy weight will have gent, and gluttonous (DeJong 1980; its own direct effect on health problems. Maddox et al. 1968). These findings The general effects of excess weight on demonstrated the prejudiced manner health have been enumerated at in which obese people are treated, but length (Andreyeva, Sturm, and Ringel as Carr and Friedman (2005) note, it 2004; Ferraro and Kelley-Moore is important to also consider whether 2003), and thorough reviews of the lit- obese people perceive that they have erature can be foundelsewhere(e.g., been mistreated. This subjective attri- Houston, Nicklas, and Zizza 2009; bution of discrimination to one’s Kopelman 2007). In brief, excess weight is a key component of the stig- weight has a pervasive effect on matization process and helps explain a host of bodily systems, affecting why a stigmatized trait would affect metabolism, endocrinology, respira- someone’s life chances (Carr and tion, and musculoskeletal integrity, Friedman 2005). among other aspects of health.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 79

Perhaps not surprisingly, obese peo- physiological processes linked to adi- pletendtoratetheirhealthmorepes- pose tissue. We therefore anticipate simistically than do normal weight that severe obesity raises the risk of individuals, and this association perceived weight discrimination and persists even when accounting for the identification as an overweight person presence of disease and functional abil- and that the effect of stigma on health ity (Ferraro and Yu 1995; Goldman, associated with the stress of stigma Glei, and Chang 2004; Okosun et al. wouldbeobservedmostacutelyin 2001). Ferraro and Yu (1995) suggest this group. that obese persons’ cognizance of the health risks associated with their PERCEIVED DISCRIMINATION, body weight influences self-health STIGMA, AND HEALTH ratings above and beyond what is cap- tured with objective health measures. Moving beyond the issue of obesity per One of the important considerations se, this paper also considers how stigma of studying weight’s effect on health is may get ‘‘under the skin’’ (McEwen the variability across levels of body 1998) and affects health. This position mass. As the prevalence of obesity rises is concordant with a long line of in a society, it is possible that the stigma research on the stress process, which associated with obesity will not be as argues that disadvantaged position in acutely felt. Over a third of adults in status hierarchies produce ill health the United States, for instance, are effects (Thoits 1995). This issue has obese by medical standards and over typically been studied with reference another third are considered overweight to the ascribed status of race. Self-rated (Flegal et al. 2010). Though American health, chronic conditions, disability, society has grown increasingly rotund depression, and blood pressure are over the past four decades, there is little affected by perceived racial discrimina- evidence to suggest that weight-based tion (Schnittker and McLeod 2005). The discrimination is disappearing—partic- stress consequences of perceived dis- ularly for those at the heaviest end of crimination extend to even the antici- the weight spectrum. pation of discrimination, suggesting Severe obesity may be the new that racial minorities often live in threshold to distinguish excess weight a chronic state of physiological arousal and where stigma’s effect may have and ‘‘heightened vigilance’’ (Williams the most import. The conventional cut- and Neighbors 2001). point for Class I obesity is a body mass A growing body of recent research index (BMI) of 30 to 34, whereas Class supports the proposition that internal- II obesity is a BMI from 35 to 40 and izing weight-related stigma has nega- Class III is 40 and above. Past studies tive consequences. Muennig (2008) indicate that health complications are has recently advanced an intriguing increasingly heightened for severely proposition—that part of heavy obese individuals, those persons with weight’s effect on morbidity owes to a BMI categorized as Class II or III the stressful burden of enduring a stig- (Andreyeva et al. 2004). One expects matized position in ‘‘body conscious’’ that severe obesity raises health societies (Crossley 2004). Perceived risks, but this could be due in part to mistreatment on the basis of one’s the social processes associated with weight explains, for instance, the interactions rather than just the higher prevalence of psychological

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 80 Social Psychology Quarterly 74(1) distress and lower self-acceptance anti-fat sentiments shapes their health among severely obese people (Carr behaviors. Puhl and Brownell (2006) and Friedman 2005; Carr, Friedman, report that in a sample of over 2,000 and Jaffe 2007).3 Weight is an interest- overweight and obese women, 79 per- ing contrast to race in that the former cent indicated that they coped with is a primarily achieved status, whereas weight stigma by eating more food, and the latter is ascribed and practically 75 percent refused to diet as a response immutable. Although both statuses to weight stigma. Recent findings even are related to stratification processes, suggest that weight-related stress and ‘‘fatness’’ offers far less leverage as dissatisfaction explain a large portion a protective identity to buffer the of obesity’s effect on health (Muennig stress of perceived mistreatment et al. 2008). We therefore expect that if (Puhl and Brownell 2001). perceived discrimination on the basis of body weight poses additional health risks to obese people, its effect will be THE ROLE OF PERCEIVED reflected in people’s interpretation of WEIGHT STATUS their weight status. That is, internalized If perceived discrimination adds extra perceptions about one’s weight status challenges to the already existing will account for the damaging effects of health threats posed by obesity, the perceived discrimination. process of stigma internalization may explain the exacerbation of this effect. Though to our knowledge this has not HYPOTHESES: been empirically tested, past research CONSEQUENCES OF makes such a thesis reasonable. Part of PERCEIVED WEIGHT the explanation for why severely obese DISCRIMINATION people show lower self-acceptance comes We generated several hypotheses for from evidence that obese people show this analysis, and they are divided levels of anti-fat bias that are similar into expectations focused on identity to their leaner counterparts, as demon- or health. Drawing on Goffman’s per- strated by experimental evidence indi- spective of stigma, we view identity as cating that obese people themselves har- a social process, influenced by one’s bor implicit devaluations of ‘‘fat people’’ perceptions of how he or she is viewed (Wang, Brownell, and Wadden 2004). by others, especially when these per- There is also evidence to suggest ceptions involve pejorative aspects of that obese people’s internalization of the self. Because stigma involves a reflexive response to the behavior of 3It is important to emphasize that perceived discrimination is fundamentally about people’s others, our first hypothesis is: appraisal of their situation; two people may inter- pret the same situation very differently, one Hypothesis 1: Perceived weight discrimi- attributing ill will to the circumstances while nation increases the likelihood that the other considering the events benign. a person self-identifies as being Presumably, the former individual would suffer overweight. worse consequences because of the stress induced by her negative appraisals of the situation We view the hypothesized influence (Lazarus and Folkman 1984). Indeed, studies of of perceived weight discrimination on perceived racial discrimination show that such appraisals induce a health-compromising stress an overweight identity as above and process (Sellers et al. 2003). beyond the effects of actual body weight.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 81

Based on prior research, we also expect METHODS that white adults and women will be Data are drawn from two waves of more likely than black adults and men, phone and self-administered question- respectively, to see themselves as naire data from the National Survey belonging to a heavier body weight cate- of Midlife Development in the United gory (Schieman, Pudrovska, and Eccles States (MIDUS). Initial data were col- 2007). We predict that perceived weight lected from 1995 to 1996 by the discrimination will influence weight MacArthur Foundation’s Network on identification net of these factors as well. Successful Midlife Development. The When perceived discrimination by survey first used random digit-dialing others affects how people identify their to obtain a sampling frame of all weight status, we also anticipate that English-speaking, non-institutionalized health will be compromised. This expec- adults aged 25 to 74 in the contiguous tation guides our second and third 48 states. The investigators then used hypotheses. disproportionate stratified sampling to oversample males between 65 and 74. Hypothesis 2: Perceived weight discrimination increases the likelihood of The response rate from these initial tele- health problems, exacerbating the phone interviews was 70 percent. The effect due to body weight alone. final stage included a questionnaire mailed to those who participated in the The second hypothesis provides a gen- telephone interview, yielding an 86.6 eral expectation drawn from the extant percent response rate. Thus, the overall literature on the effects of perceived dis- response rate for Wave I was 61 percent crimination on health (Gee 2002; (.70 3 .87 = .61), producing a total sam- Krieger 1999; Williams et al. 2003) and ple of 3,034 participants who completed the damning effects of stigma on life both the telephone and mail interview. chances (Link and Phelan 2001). Respondents were then recontacted Although most previous studies identify to secure their participation for Wave the health consequences of excess II (2005). Of the complete Wave I sam- weight as driven largely by physiological ple, 2,103 individuals (69 percent) were processes, the significance of this followed up on the telephone. Cases hypothesis is to determine if unfair with missing data on variables collected treatment aggravates the health prob- at baseline were dropped from the anal- lems of persons with excess weight. yses, leaving a final study sample of 1,856 for the majority of analyses.4 For Hypothesis 3: Weight perceptions mediate one set of analyses, however, the sample the exacerbating effect of perceived shrank to 1,560 because questions about weight discrimination on health. functional disability at Wave II were in the mailed questionnaire and some Our final hypothesis specifies that when weight discrimination is internalized and shapes weight identification, 4Of the missing Wave I variables, BMI data such identification explains the exacer- were the most prevalent (96 missing cases). bating effect that perceived discrimina- BMI was gathered from respondents’ height and tion has on health. When testing weight reports. It is worth noting that those Hypotheses 2 and 3, we also predict who did not provide their height and weight were very similar to the study sample in terms that severe obesity will pose the greatest of other health status variables and in regard to health risk. reports of discrimination.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 82 Social Psychology Quarterly 74(1) respondents did not return this ques- Mediating Variables tionnaire. Post-stratification weights Perceived weight status is based on are available for the second wave of a question in which respondents were the data and are used in all multivariate asked, ‘‘Which of the following do you analyses in order to draw generalizeable consider yourself? (1) very overweight, conclusions to the adult U.S. population. (2) somewhat overweight, (3) about the One of the important considerations right weight, (4) somewhat under- for using more than one wave of survey weight, or (5) very underweight.’’ data is the potential biasing effect due Given that small percentages of to attrition (Winship and Mare 1992). respondents identified as categories 4 This is especially a concern because (2.95 percent) and 5 (.20 percent), we the key study variables measure collapsed and re-ordered so that very health status, which is clearly related overweight was the high category, some- to the chance of attrition, particularly what overweight was the middle cate- from mortality. Because attrition may gory, and not overweight was the lowest lead to specification error and bias in category. Each category was coded as theresults,wefollowedtheHeckman a dummy variable; for multivariate (1979) method of correcting for nonre- analyses, the not-overweight category sponse bias. serves as the reference group.

Outcome Variables Independent Variables Health status is operationalized with Body weight was measured by asking two variables, each measured at Wave respondents to report their height and I and Wave II. Our first measure is weight. Self-reported weights are functional disability, which was widely acknowledged as valid instru- assessed with nine questions about ments, but provide slight underesti- how much a respondent is limited in mates of weight distribution extremes activities such as transporting grocer- (Bowman and DeLucia 1992). Body ies and walking over a mile (a = .87). mass index (BMI) was calculated by Disability is an important health out- the formula kilograms/meters2, and come for obesity studies because of the two binary variables were created to ways that excess weight limits mobility differentiate between people of Class I and the negative implications this has obesity and severe obesity (Class II and for social interaction (Ferraro and Class III; see National Heart, Lung, Kelly-Moore 2003). Our second health and Blood Institute 1998). We specified indicator is an indicator of self-rated models with the categorical or continu- health ranging from 1 (poor) to 5 (excel- ous measure of body weight based on lent). This global evaluative health the aims of each analysis. measure is a strong predictor of mortal- To measure perceived weight discrim- ity and morbidity and has long been ination, we used a set of questions about used in the social psychology of health appraisals of discriminatory experiences. (Seeman, Seeman, and Sayles 1985). Thisunitofthesurveyfirstasked For both health outcome variables, we respondents if they have experienced created a change score between the any interpersonal offenses, including two waves (i.e., Wave II disability– instances in which (1) people act as if Wave I disability; Wave II self-rated you are inferior; (2) people act as health–Wave I self-rated health). if you are not smart; (3) people act as

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 83 if they are afraid of you; (4) you are resulted in a series of dummy varia- treated with less courtesy than others; bles. The omitted reference category (5) you are treated with less respect in regression models is non-obese than others; (6) you receive poor respondents who did not perceive services in stores/restaurants; (7) peo- discrimination.6 ple act as if you are dishonest; (8) you Analyses include controls for a num- are called names or insulted; and (9) ber of variables associated with both you are threatened or harassed. Next, self-rated health and body weight, all of respondents were asked if they had which were measured at Wave I. First, ever faced discrimination in a host of for health-related indicators, we incorpo- social settings (e.g., workplace, school). rate a summary of chronicconditionsat Respondents who answered affirma- Wave I (e.g., heart trouble, thyroid prob- tively to any of these scenarios were lems, arthritis) as well as self-rated then asked to identify the reason for health at Wave I. Another health-related the perceived discrimination, includ- indicator includes whether the respon- ing race, age, gender, ethnicity, reli- dent lives a sedentary lifestyle (little or gion, disability, sexuality, weight/ no exercise). Mental health and well- height, and other aspects of physical being was assessed with a six-item scale appearance. Respondents could select measuring negative affect (a = .87), and more than one reason for the perceived we used an averaged score. Examples of discrimination. About seven percent of the items include feeling ‘‘restless or the sample reported any type of per- fidgety’’ and ‘‘hopeless’’ (mood was ceived weight discrimination and we assessed for the past 30 days). Finally, coded this as a binary variable.5 we include a binary variable for current Testing Hypothesis 2 involved an smoking status. interaction between body weight Second, the demographic characteris- and perceived weight discrimination. tics of race and gender are coded as Although we tested product terms from binary variables (1 for black; 1 for two binary variables in preliminary female), and age is a continuous vari- analyses, we elected to create separate able. An age-squared term was explored, binary variables to represent the but since it did not improve model fit for cross-classification of the two variables the models presented herein, it was (because of the ease of interpretation). removed from the final analyses. Two In other words, Class I and severely variables tapping socioeconomic status obese respondents were differentiated are incorporated. Education is an bywhetherornottheyperceivedany discrimination based on weight. This 6Because only two underweight respondents perceived weight discrimination, we created only one dummy variable for underweight, not 5Whereas the wording of the question actually differentiating on the basis of discrimination. probes weight or height discrimination, we exam- Thus, the full set of nine binary variables reflect- ined the percent of persons in each BMI classifi- ing the weight class/discrimination categories cation who reported this type of discrimination. are: (a) underweight, (b) normal weight and no The pattern is highly indicative of a weight-cen- discrimination, (c) normal weight and discrimi- tered approach to understanding the survey nated against, (d) overweight and no discrimina- question; among the respondents, 33 percent of tion, (e) overweight and discriminated against, the severely obese, 18 percent of the obese, 4 per- (f) Class I obese and no discrimination, (g) Class cent of overweight, 2 percent of normal weight, I obese and discriminated against, (h) severely and 5 percent of underweight report this type of obese and no discrimination, and (i) severely discrimination. obese and discriminated against.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 84 Social Psychology Quarterly 74(1) ordinal variable ranging from 1 (some Analytic Strategy grade school) to 12 (doctoral or profes- After presenting descriptive sample sional degree). Household income is statistics and showing basic compari- a logged continuous variable (which in sons in our key variables across body raw form has a mean of $58,427 with weight classifications, we proceed in a standard deviation of $49,188). two main analytic stages. Our overall approach is to examine weight identifi- Nonresponse Selection and Missing cation as a mediating factor in the Data relationship between perceived weight discrimination, obesity, and health To apply Heckman’s (1979) method of status. To that end, we will first show correcting for potential nonresponse the initial path in the proposed bias in longitudinal data, we began by process—the association of perceived first estimating a probit model discrimination and self-perceptions of predicting the likelihood of Wave II weight status. Although weight percep- response, using a variety of demographic tions are an ordered categorical vari- and psychosocial variables as predictors. able, preliminary analyses using the The next step was to calculate a nonre- likelihood ratio test of proportionality sponse hazard score (l), based on the of odds across response categories indi- inverse Mills ratio of the function cated that using an ordinal logistic derived from the probit model. This regression model would violate the score is considered the hazard of nonre- model’s parallel lines assumption sponse and included as a control variable (Long and Freese 2006). Because of in regression estimates. this violation, we elected to use a multi- The second complexity of using the nomial logistic regression model that MIDUS data is that the survey consists makes no assumptions about order of of two parts (i.e., telephone interview the categories. We will present relative and mailed questionnaire). Although risk ratios, which are the change in 2,103 respondents who completed probability of being in the specified cat- Wave I were followed up at Wave II, egory versus the baseline category a total of 355 responded only to the tele- across adjacent levels of an indepen- phone interview and did not return the dent variable.7 mailed questionnaire. To preserve Wave II subjects who were followed up but did not complete the second portion 7In order to rule out a reverse explanation— of the survey, we imputed missing that weight perceptions make the perception of Wave II data on perceived weight discrimination more likely—we also estimated status, which was measured in the logit models predicting the probability of Wave mailed questionnaire. Imputation for II perceived discrimination. Weight perceptions missing data consisted of estimating at Wave I, however, did not have a significant association with perceptions of discrimination. equations for the missing variables This finding is distinct from those reported in with demographic, psychosocial, and the literature on racial identity and perceived dis- health information and using predicted crimination, suggesting that centrality of racial scores for missing values. In sensitivity identity increases the likelihood of reporting dis- analyses, we also replicated our models criminatory behavior (Sellers et al. 2003). We tested but could not find any evidence to support without imputation and with multiple the notion that weight perceptions increase per- imputation, and the conclusions were ceived weight discrimination (or weight, for that similar to those presented here. matter).

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 85

If perceived weight discrimination reported that they were ‘‘somewhat affects weight perceptions, we can overweight,’’ though a sizeable minor- proceed to test the second half of the ity reported being ‘‘very overweight’’ proposed process. The second portion or not being overweight. In addition, of the analyses therefore will examine about a quarter of the sample was obese the effects of perceived weight discrim- (BMI 30). ination on disability and self- Table 2 highlights the differences in rated health with and without includ- perceived weight discrimination across ing perceptions of weight status. An weight classifications. Whereas fewer initial model will show the simple, than 5 percent of non-obese respondents unadjusted relationship of weight clas- perceived such discrimination, almost sification on health change to ascer- 11 percent of Class I obese respondents tain that a basic association exists. and 33 percent of severely obese These analyses will utilize OLS respondents did. regression. After determining that severity of Themodelswewillestimateusetwo obesity is related to the likelihood that waves of panel data, with independent one will perceive discrimination due to and control variables measured at his or her weight, the importance of sep- Wave I predicting change in the out- arately examining Class I obese and come variables between waves. It is severely obese respondents becomes advantageous to use longitudinal data apparent. Not only may severe obesity for our research question, because affect perceived discrimination, but as using a lagged measure of body weight shown in Table 3 it also affects weight and perceived discrimination on health perceptions. The bottom several rows reduces the risk of problems associated of Table 3 confirm this expectation, as with potential reverse causality. That the likelihood of reporting being very is, with cross-sectional data, it would overweight is highest for severely obese be more difficult to rule out that poor respondents and less likely for Class I health is actually leading to perceived obese respondents. Unsurprisingly, weight discrimination. Rather, our non-obese respondents are much less approach is to use body weight and likely to feel very overweight; none of perceived weight discrimination at the underweight participants and fewer Wave I to predict change in health than 2 percent of all normal weight fromWaveItoWaveII. respondents reported such a status. More relevant for our purposes, however, is the effect of perceived RESULTS weight discrimination on weight percep- As shown in Table 1, the change score tions across the groups. In the middle for functional disability between waves rows of Table 3, for instance, we observe reveals that the average level of disabil- that 7 percent of overweight subjects ity was slightly higher at Wave II. Most whodidnotperceiveweightdiscrimina- respondents increased at least slightly tion felt very overweight, but this figure in functional disability, though some was elevated to 20 percent for over- participants had lower levels of Wave weight respondents who perceived II disability than they did at Wave I. discrimination. Likewise, although only As Table 1 also indicates, self-rated about a quarter of Class I obese health declined slightly on average respondents felt very overweight if between waves. Most respondents also they had not perceived weight

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 86 Social Psychology Quarterly 74(1)

Table 1. Descriptive Sample Statistics for Variables Used in the Full Study Sample, MIDUS (n = 1,856)

Range Mean Standard Deviation

Dependent Variables (WII) Functional disability change (WI–WII)a 23–2.78 .27 .65 Self-rated health change (WI–WII) 24–3 2.04 .93 Mediating Variables (WII) Perceived weight status Feel not overweight 0–1 30.60% Feel somewhat overweight 0–1 55.50% Feel very overweight 0–1 13.90% Independent Variables Weight discrimination 0–1 6.84% Body weight BMI (kg/m2) 9–61 26.69 5.30 Underweight (BMI \ 18.5) 0–1 2.10% Normal (BMI 18.5–24.9) 0–1 35.18% Overweight (BMI 25–29.9) 0–1 38.85% Class I obese (BMI 30–34.9) 0–1 16.11% Severe obese (BMI 351) 0–1 7.76% Control Variables Health status Self-rated health (WI) 1–5 3.57 .96 Count of chronic conditions 0–32 2.57 2.60 Sedentary lifestyle 0–1 1.24% Negative affect 1–5 1.55 .63 Functional disability (WI) 1–4 1.43 .64 Smoking 0–1 20.42% Demographic controls Age 20–74 46.56 12.69 Black 0–1 4.96% — Female 0–1 51.62% — Education 1–12 7.02 2.44 Household income (ln) 6.21–12.21 10.67 .85

Note: an = 1,560 for Functional disability change between Wave I and Wave II discrimination, the likelihood of feel- Table 2. Prevalence of Perceived Weight ingveryoverweightnearlytripled(67 Discrimination by Weight Classification, MIDUS (n = 1,856) percent) if they had perceived weight discrimination. In fact, Class I obese Study Faced weight respondents who perceived discrimina- sample discrimination* tion were about as likely to see them- Underweight 2% (n = 39) 5% (n =2) selves as very overweight as were Normal 35% (n = 653) 2% (n = 14) severely obese respondents who had weight not perceived discrimination. Even Overweight 39% (n = 721) 4% (n = 30) those in the most extreme BMI category, Class I obese 16% (n = 299) 11% (n = 81) whose self-weight evaluations may be Severe obese 8% (n = 144) 33% (n = 48) thought to be impervious to others’ Notes: *Comparison of perceived weight behaviors, were 12 percent more likely discrimination prevalence across five weight to see themselves as very overweight if classes, p \ .001 (x2= 197.71, 4 df).

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 87

Table 3. Perceived Weight Status by Weight Classification and Perceived Weight Discrimination, MIDUS (n = 1,856)

Do not feel Feel somewhat Feel very overweight overweight overweight

Total Sample 31% (n = 568) 56% (n = 1030) 14% (n = 258) Underweight No weight discrimination 86% (n = 32) 14% (n =5) 0%(n =0) Faced weight discrimination 50% (n = 1) 50% (n =1) 0%(n =0) Normal No weight discrimination 58% (n = 369) 40% (n = 255) 2% (n = 15) Faced weight discrimination 43% (n = 6) 57% (n =8) 0%(n =0) Overweight* No weight discrimination 20% (n = 142) 73% (n = 501) 7% (n = 48) Faced weight discrimination 7% (n = 2) 73% (n = 22) 20% (n =6) Class I obese** No weight discrimination 5% (n = 16) 69% (n = 214) 26% (n = 132) Faced weight discrimination 0% (n = 0) 33% (n = 24) 67% (n = 57) Severe obese No weight discrimination 3% (n = 3) 32% (n = 31) 65% (n = 62) Faced weight discrimination 0% (n = 0) 27% (n = 13) 73% (n = 35)

Note: *Comparison of perceived weight status between no weight discrimination and perceived weight discrimination, p \ .01 (x2 = 9.32, 2 df). **Comparison of perceived weight status between no weight discrimination and perceived weight discrimination, p \ .001 (x2 = 32.13, 2 df). they perceived weight discrimination treatment by others shape weight per- (73 percent versus 65 percent). ceptions above and beyond the effects Table 4 builds on these findings, of one’s weight category. To be sure, showing the results of a multinomial the relative risk ratios of feeling very logistic regression model predicting per- overweight vary according to one’s ceived weight. Because all obese sub- actual weight status, as each unit jects who perceived discrimination saw increase in BMI is associated with themselves as at least somewhat over- a 1.89 increase in the likelihood of weight, we could not estimate the reporting that one feels ‘‘very over- three-response category model with the weight’’ compared to ‘‘not overweight.’’ obesity variables coded as they were in But the chief finding is that perceived Table 3. Rather, we hold BMI constant, discrimination contributes to one’s iden- add a BMI-squared term to capture tified weight status (Relative Risk Ratio potential non-linear relationships, and [RRR] = 2.58 for perceptions of feeling examine whether perceived weight dis- somewhat overweight versus not crimination has any added effect on overweight; RRR = 4.08 for perceptions weight perceptions, controlling for other of feeling very overweight versus not relevant variables.8 The significant overweight). This finding confirms effects of perceived weight discrimina- Hypothesis 1, and the findings for the tion show that appraisals of unfair other variables are largely consistent with recent research on the predictors 8Adding the BMI-squared term significantly of weight perceptions (Schieman et al. improved model (likelihood ratio test, p \ .01). 2007). Specifically, women and those of

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 88 Social Psychology Quarterly 74(1)

Table 4. Multinomial Logistic Regression Predicting Perceived Weight Status, MIDUS (n = 1,856)

Feel Somewhat Feel Very Overweight vs. Overweight vs. Not Overweight Not Overweight RRR, 95% CI p RRR, 95% CI p

Independent Variables BMI 1.55 (0.79–3.02) ns 1.89 (0.92–3.88) \.10 BMI2 1.00 (0.98–1.01) ns 1.00 (0.98–1.01) ns Faced weight discrimination 2.58 (1.02–6.52) \.05 4.08 (1.35–12.34) \.05 Perceived weight status (WI) Feel somewhat overweight 7.61 (5.17–11.01) \.001 12.67 (4.86–33.05) \.001 Feel very overweight 2.49 (0.93–6.68) \.10 39.46 (9.84–158.29) \.001 Control Variables Health status Self-rated health (WI) 0.96 (0.78–1.19) ns 0.99 (0.69–1.42) ns Count of chronic conditions (WI) 1.08 (0.99–1.19) \.10 1.05 (0.93–1.19) ns Negative affect 0.75 (0.56–1.01) \.10 0.89 (0.54–1.48) ns Sedentary lifestyle 0.25 (0.05–1.26) \.10 0.14 (0.01–3.42) ns Functional disability 1.11 (0.80–1.53) ns 1.58 (1.01–2.46) \.05 Smoking 1.03 (0.66–1.59) ns 1.00 (0.50–1.99) ns Demographic factors Age 0.99 (0.97–1.00) ns 0.97 (0.94–1.00) \.05 Black 1.03 (0.36–2.88) ns 0.78 (0.17–3.53) ns Female 3.23 (1.96–5.32) \.001 9.32 (4.53–20.89) \.001 Education 1.02 (0.92–1.12) ns 0.96 (0.82–1.12) ns Household income (ln) 1.23 (1.02–1.48) \.01 1.88 (1.30–2.72) \.001 Nonresponse hazard 3.51 (0.52–24.04) ns 9.75 (0.39–242.19) ns 22 Log-likelihood 21082.54 Pseudo R2 .40

Note: RRR = Relative risk ratio. CI = Confidence interval. Tests are two-tailed. higher household incomes were more explain this relationship. The first three likely to report being somewhat over- columns of Table 5 show the results of weight or very overweight. The gender three models predicting change in func- effect was particularly strong (e.g., tional disability. Model I shows the sim- RRR = 9.32 for feel very overweight ver- ple effect of weight class on functional sus not overweight). We did not, how- disability change. Models II and III dis- ever, observe a difference between black aggregate weight classes into those who and non-black respondents. All results perceive discrimination and those who presented in Table 4 control for baseline do not. The omitted reference group for weight perceptions, thus reflecting any weight class and discrimination is nor- change in weight perceptions between mal weight subjects who did not per- Wave I and Wave II. ceive weight discrimination. (Because The remainder of the analyses consid- only two underweight respondents per- ers whether perceived weight discrimi- ceived weight discrimination, we do not nation exacerbates health decline over differentiate such participants on the a decade and to what extent the weight basis of weight discrimination.) Model identification processes identified above II is specified to first ask whether

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 Table 5. OLS Regression Predicting Functional Disability Change (n = 1,560) and Self-Rated Health Change (n = 1,856), MIDUS

Functional Disability Self-Rated Health Model I Model II Model III Model I Model II Model III Downloaded from b, S.E. p b, S.E. p b, S.E. p b, S.E. p b, S.E. p b, S.E. p

Independent Variables spq.sagepub.com Underweight 2.03 (.09) ns .07 (.09) ns .10 (.09) ns .17 (.13) ns .12 (.12) ns .08 (.12) ns Normal Faced weight discrimination 2.08 (.10) ns 2.07 (.09) ns 2.01 (.19) ns 2.02 (.19) ns Overweight 2.01(.04) ns 2.09 (.05) \.10 atASA-American SociologicalAssociation onMarch4,2011 No weight discrimination 2.02 (.04) ns 2.06 (.04) ns 2.04 (.05) ns .00 (.06) ns Faced weight discrimination .11 (.12) ns .07 (.13) ns 2.11 (.11) ns 2.04 (.11) ns Class I obese .26 (.06) \.001 8 9 2.23 (.07) \.001 8 9 No weight discrimination > .07 (.07)> ns .01 (.07) ns >2.10 (.08)> ns .01 (.09) ns > > > > Faced weight discrimination < .32 (.16)=\.05 .20 (.16) ns <2.24 (.15)= ns 2.12 (.15) ns Severe obese .26 (.08) \.001 > > 2.39 (.09) \.001 > > > > > > No weight discrimination :> .01 (.10);> ns 2.10 (.11) ns :>2.23 (.12);>\.05 2.03 (.12) ns Faced weight discrimination .38 (.13) \.01 .27 (.13) \.05 2.33 (.13) \.01 2.11 (.14) ns Perceived weight status Feel somewhat overweight .05 (.04) ns .05 (.05) ns Feel very overweight .20 (.07) \.01 2.36 (.09) \.001 Control Variables Health status Self-rated health (WI) 2.06 (.02) \.01 2.06 (.02) \.01 2.45 (.02) \.001 2.60 (.03) \.001 2.61 (.03) \.001 Functional disability (WI) 2.22 (.04) \.001 2.36 (.05) \.001 2.37 (.05) \.001 2.21 (.05) \.001 2.19 (.05) \.001 Negative affect .03 (.03) ns .03 (.03) ns 2.08 (.04) \.05 2.07 (.04) \.10 Count of chronic conditions .04 (.01) \.001 .04 (.01) \.001 2.01 (.01) \.10 2.01 (.01) ns (continued) 89 90 Downloaded from Table 5. (continued)

Functional Disability Self-Rated Health spq.sagepub.com Model I Model II Model III Model I Model II Model III b, S.E. p b, S.E. p b, S.E. p b, S.E. p b, S.E. p b, S.E. p

atASA-American SociologicalAssociation onMarch4,2011 Sedentary lifestyle 2.32 (.18) \.10 2.31 (.16) \.05 .33 (.27) ns .31 (.24) ns Smoking .07 (.05) ns .07 (.05) ns 2.21 (.06) \.001 2.21 (.06) \.001 Demographic factors Age .01 (.00) \.05 .01 (.00) \.05 2.01 (.00) \.05 2.01 (.00) \.05 Black .05 (.09) ns .05 (.09) ns 2.35 (.12) \.01 2.35 (.12) \.01 Female 2.08 (.07) ns 2.11 (.07) ns .09 (.08) ns .14 (.08) \.10 Education 2.04 (.01) \.001 2.04 (.01) \.001 .05 (.01) \.001 .05 (.01) \.001 Household income (ln) 2.01 (.02) ns 2.02 (.02) ns .04 (.03) ns .05 (.03) \.10 Nonresponse hazard 2.98 (.50) \.10 2.96 (.51) \.10 .40 (.52) ns .37 (.51) ns R2 .06 .20 .21 .20 .29 .30 N 1560 1560 1560 1856 1856 1856

Note: OR = Odds ratio. CI = Confidence interval. B = Unstandardized coefficient. S.E. = Standard error. Tests are two-tailed. Normal weight and no perceived weight discrimination is the reference group. Brackets represent coefficients that were tested for equality with joint Wald tests (no weight discrimination coefficients tested for equality with perceived weight discrimination coefficients). An asterisk denotes significant difference between weight class and no weight discrimination and weight class with perceived weight discrimination. The Stigma of Obesity 91 perceived weight discrimination aggra- (Model I), results show that Class I vates weight-related increases in dis- obese and severely obese respondents ability, and Model III includes weight have worse self-rated health a decade perceptions as mediating variables. later than do normal weight persons. Model I indicates that obese and Just as in the functional disability mod- severely obese respondents fared the els, the health effects of obesity appear worst in regard to functional disability more severe for those who perceived between survey waves, as would be weight discrimination. Results from expected. Model II indicate that there was not Turning to Model II, the unstandard- a statistically significant effect of Class ized change coefficient was largest for I obesity—whether accompanied by per- those who were severely obese and perceived discrimination or not—on health ceived weight discrimination (b =.38, decline. Similar to the models’ predic- p \ .01). Interestingly, those who were tion of change in disability, however, Class I obese and perceived discrimina- the consequence of severe obesity cou- tion faced greater increases in disability pled with perceived weight discrimina- (b =.32,p \ .05) than severely obese tion produced worse health declines people who did not perceive discrimina- than did severe obesity without weight tion (b = .01, nonsignificant). Joint discrimination, as reflected by the size Wald tests for a comparison of effect of the coefficients (b = 2.33 versus sizes reveal that the coefficients for 2.23). However, the difference in the weight class and discrimination are size of the coefficients fails to reach sig- greater than the weight classes and no nificance when comparing effect sizes discrimination (F =4.73,p \ .01) with the Wald tests, thus failing to These findings support Hypothesis 2. directly support Hypothesis 2. Not surprisingly, those with higher In support of the idea that weight self-rated health at Wave I did not perceptions are consequential for health increase in disability between waves, decline, however, the findings from whereas a higher number of chronic Model III show that the influence of health conditions were associated with weight status on self-rated health increasing disability. among severely obese persons becomes When including the weight percep- nonsignificant when weight perceptions tions variable in Model III, the coeffi- are included as mediating variables, cient for severely obese and perceived though other variables significant in weight discrimination was attenuated Model II retain their significance. somewhat (b = .27, down from .38 in Thus, there is mediation of actual Model I) and the effects for Class I obe- weight status (accompanied or not by sity became nonsignificant, supporting perceived discrimination) by heavier Hypothesis 3. Though the effect of feel- weight identities, which are shaped by ing only somewhat overweight was non- perceived discrimination. Given the significant, respondents who felt very absence of the hypothesized exacerba- overweight experienced an increase in tion effect (Hypothesis 2), Hypothesis 3 disability between Wave I and Wave II cannot be fully confirmed; nevertheless, (b =.20,p \ .01). the results are consistent with its The three rightmost columns in Table essence. The perception that one is 5 undertake a parallel analysis but with only somewhat overweight is not signif- change in self-rated health as the out- icantly associated with decreasing self- come variable. In the fourth column rated health, but perceiving one’s self

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 92 Social Psychology Quarterly 74(1) as being extremely overweight is related interested us here for several reasons. to a .36 unit decrease in self-rated For one, a growing body of research health relative to perceiving that one is identifies that perceived discrimination not overweight. poses a threat to health (e.g., Williams et al. 2003), but weight-based discrimination has been largely left out of this DISCUSSION conversation. In addition, scores of In modern, body-conscious societies, research articles have documented how heavy weight—obesity in particular— excess weight is harmful to the body, may imply some level of reprehensibil- but the exacerbating contribution of ity (Crossley 2004), or what Goffman social factors to the health risks of obe- (1963) referred to as a ‘‘deeply discred- sity are not clearly understood. iting’’ trait producing a ‘‘spoiled iden- When we differentiate between obese tity.’’ National survey data indicate persons who have perceived weight dis- that obese people are at risk of per- crimination and those who do not, a clear ceived maltreatment (Carr and picture emerges: perceived discrimina- Friedman 2005), but the health conse- tion aggravates problems with mobility. quences of such discrimination have Weight-based perceived discrimination heretofore not been examined. With nearly evened the differences between longitudinal data, this study examines Class I and severe obesity in terms of the antecedents of perceived weight functional disability; individuals in status and whether obesity stigma either weight category had increases in exacerbates health problems, net of disability over 10 years, and Class I baseline health and demographic fac- obese adults who perceived discrimina- tors. We find that obesity stigma affects tion fared worse than severely obese health, particularly in regard to change adults who did not perceive discrimina- in functional disability, and that weight tion. These findings challenge a medical- perceptions are key to understanding ized absolutism positing that severe lev- this relationship. els of weight pose threats only through Our main findings can be summa- direct physiological means. Our results rized around two main themes. First, reveal that the social processes of per- we showed that people are likely to perceived weight discrimination are respon- ceive themselves as heavier if they have sible, at least in part, for the deleterious perceived weight discrimination, an effects of severe obesity on health. association that persisted even when The findings on self-rated health are controlling actual weight status. Body somewhat less clear. Although the lin- weight has the potential to be a discred- ear regression coefficients were larger iting trait (Cahnman 1968; Carr and for severely obese people if they per- Friedman 2006; DeJong 1980; Maddox ceived discrimination, the difference et al. 1968; Puhl and Brownell 2001), was not statistically significant, perhaps and mistreatment on the basis of that because of the relatively small number characteristic is an important force on of severely obese people in the sample. one’s self-concept. As we tie together these reports of Drawing on Goffman (1963), we discrimination with the more general expected that an identity as an over- concept of stigma and the far-extending weight person would pose constraints damage that stigmatized traits pose for on life chances, and this organizes our life chances (Phelan, Link, and Dovidio second set of findings. Health outcomes 2008), the results become more telling:

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 93 when perceived weight status is exam- Interestingly, this picture of per- ined alongside actual weight status ceived discrimination, identity, and (accompanied or not by perceived dis- health runs counter to some recent find- crimination), the latter no longer has ings involving race. Neblett et al. (2004), a direct effect on self-rated health for instance, find that racial discrimina- declines. For changes in disability, the tion is less detrimental for health among effect of being Class I obese and perceiv- people with salient racial identities. In ing discrimination is explained by light of other studies, however, this is weight perceptions brought on by per- rather unsurprising. Whereas strong ceived discrimination. The effect of racial identity is generally reported as being severely obese and perceiving dis- a buffer to stress and health threats crimination is lessened, though not (Mossakowski 2003; Sellers et al. entirely explained by weight perceptions 2003), heavy weight is pervasively con- (effect size reduced by 29 percent). sidered a negative aspect of self-concept, Our findings are potentially impor- and a tiny proportion of heavy people tant because they suggest that the sense embrace the ‘‘fat’’ identity (LeBesco of being marginalized because of one’s 2004; Puhl and Brownell 2001). If people weight can actually contribute to tend not to rally around a shared sense steeper health declines. In other words, of feeling heavy, then having a ‘‘fat iden- social factors are implicated in health tity’’ would offer little consolation in the problems associated with obesity along- midst of perceived wrongdoing. The side the widely recognized physiologic sense of camaraderie attached to ethnic- causes. This adds to the growing body ity, on the other hand, can be empower- of literature showing how social factors ing and more effectively stifle the insults get ‘‘under the skin’’ and affect disease of offenders. processes (Ferraro and Shippee 2009; Several limitations of the present Glass and McAtee 2006; Muennig et al. analysis must be kept in mind. First, 2008). because this was a longitudinal study, This study also builds on efforts to the problem of sample attrition suggests link discrimination and stigma, related that the results may be biased by the concepts that have historically devel- selective group who was successfully fol- oped in parallel literatures (Phelan lowed up 10 years after baseline. We et al. 2008). Goffman’s (1963) idea of accounted for this problem by employing stigma emphasized that prejudicial the Heckman procedure, but there is the actions of others—particularly when possibility that other selection factors enforcing the norms of desirable traits may operate. Second, the way that ques- (e.g., thinness, beauty)—effectively tions about perceived discrimination exclude one from full participation in were asked on the survey does not allow social life. Yet, at least in the case of obe- one to rule out the possibility that traits sity, a good part of what makes this in addition to obesity may have contrib- exclusionary action harmful is its influ- uted to the maltreatment. Respondents ence on self-perceptions. Discrimina- were able to identify numerous reasons tory actions that are not perceived as for their discrimination, and so gender, such may simply be deflected, whereas race, age, or other factors could be con- perceived mistreatment understood to flated with weight, thus overestimating be related to a certain discrediting the influence of perceived weight dis- characteristic is absorbed into one’s crimination. Although we believe cau- self-concept and limits life chances. tion is warranted in interpreting these

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 94 Social Psychology Quarterly 74(1) findings, it should also be recognized a reductionistic biomedical narrative, it that only seven percent of the sample is imperative that sociologists stake reported weight-based discrimination. their ground and emphasize the irreduc- Given the skewed distribution of the ibility of social forces as the primary variable, we believe our conclusions are medium of personal fortunes (Duster quite plausible. Third, additional health 2006). Recent evidence suggests that outcomes merit attention, but many of those steeped in a worldview of reduc- the health problems implicated by obe- tionistic accounts of health often find sity and severe obesity were relatively the contribution of social scientists to rare in our data and thus proved chal- be slight or nonexistent (Albert et al. lenging to include in analyses because 2008). It is our hope that sociologists of low statistical power. Finally, the do not grow disheartened by the domi- 10-year study period was useful for nance of a biomedical research para- observing changes in both overweight digm, but continue to give attention to identity and health outcomes such as the intrinsically interpersonal and social disability, but it is a fairly long time dur- dynamics of health in contemporary ing which other important changes may society. have transpired. Future research may profit from using shorter study periods ACKNOWLEDGEMENTS to replicate or refute the processes uncovered here. We thank Dawn Alley, Ann Howell, Brian Kelly, These limitations considered, the cur- Sarah Mustillo, and Tetyana Pylypiv Shippee for rent study offers new directions for the helpful comments on earlier drafts of this paper. In addition, we gratefully acknowledge the study of obesity and health. With the insightful feedback of the SPQ editor and growing prevalence of obesity, it is reviewers. The data were made available by the becoming increasingly important to Inter-university Consortium for Political and understand the ways in which social Social Research, Ann Arbor, Michigan. Neither relationships and context exacerbate or the collector of the original data nor the Consortium bears any responsibility for the anal- minimize consequences for health. yses or interpretations presented here. We also document a large effect of gender on weight perceptions, a finding REFERENCES that is consistent with recent research (Schieman et al. 2007) and that begs Albert, Mathieu, Suzanne Laberge, Brian D. for further exploration as to whether Hodges, Glenn Regehr, and Lorelei stigma internalization processes gener- Lingard. 2008. ‘‘Biomedical Scientists’ Perception of the Social Sciences in alize between men and women. Health Research.’’ Social Science & Unfortunately our sample was too Medicine 66:2520–31. restricted in the number of severely Andreyeva, T., R. Sturm, and J. S. Ringel. obese respondents to make meaningful 2004. ‘‘Moderate and Severe Obesity Have statistical comparisons across gender Large Differences in Health Care Costs.’’ Obesity Research 12:1936–43. lines or with other important character- Bowman, Robert and Janice Delucia. 1992. istics related to body image norms, such ‘‘Accuracy of Self-Reported Weight: A as race (Carr and Friedman 2005; Meta-Analysis.’’ Behavior Therapy 23: Crosnoe et al. 2008; Schieman et al. 637–55. 2007). Cahnman, Werner J. 1968. ‘‘The Stigma of Obesity.’’ The Sociological Quarterly Finally, in an era when explanations 9:283–99. for complex processes related to well- Campbell, Mary E. and Lisa Troyer. being are increasingly dominated by 2007. ‘‘The Implications of Racial

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 95

Misclassification by Observers.’’ American of the American Medical Association Sociological Review 72:750–65. 303:235–241. Carr, Deborah and Michael A. Friedman. Gee, Gilbert C. 2002. ‘‘A Multilevel Analysis of 2005. ‘‘Is Obesity Stigmatizing? Body the Relationship between Institutional and Weight, Perceived Discrimination, and Individual Racial Discrimination and Psychological Well-Being in the United Health Status.’’ American Journal of States.’’ Journal of Health and Social Public Health 92:615–23. Behavior 46:244–59. Glass, Thomas A. and Matthew J. McAtee. Carr, Deborah and Michael A. Friedman. 2006. ‘‘Behavioral Science at the 2006. ‘‘Body Weight and the Quality Crossroads in Public Health: Extending of Interpersonal Relationships.’’ Social Horizons, Envisioning the Future.’’ Social Psychology Quarterly 69:127–49. Science & Medicine 62:1650–71. Carr, Deborah, Michael A. Friedman, and Goffman, Erving. 1963. Stigma: Notes on the Karen Jaffe. 2007. ‘‘Understanding the Management of Spoiled Identity. Relationship between Obesity and Positive Englewood Cliffs, N.J.: Prentice Hall. and Negative Affect: The Role of Goldman, Noreen, Dana A. Glei, and Ming- Psychosocial Mechanisms.’’ Body Image Cheng Chan. 2004. ‘‘The Role of Clinical 4:165–77. Risk Factors in Understanding Self- Crosnoe, Robert. 2007. ‘‘Gender, Obesity, and Rated Health.’’ Annals of Epidemiology Education.’’ Sociology of Education 14:49–57. 80:241–60. Heckman, James J. 1979. ‘‘Sample Selection Crosnoe, Robert, Kenneth Frank, and Anna Bias as a Specification Error.’’ Economet- Strassmann Mueller. 2008. ‘‘Gender, Body rica 47:153–61. Size and Social Relations in American Houston, Denise K., Barbara J. Nicklas, and High Schools.’’ Social Forces 86:1189–1216. Claire A. Zizza. 2009. ‘‘Weighty Concerns: Crossley, Nick. 2004. ‘‘Fat is a Sociological The Growing Prevalence of Obesity among Issue: Obesity Rates in Late-Modern Older Adults.’’ Journal of the American ‘Body-Conscious’ Societies.’’ Social Theory Dietetic Association 109:1886–95. and Health 2:222–53. Kopelman, Peter G. 2007. ‘‘Health Risks DeJong, William. 1980. ‘‘The Stigma of Associated with Obesity and Overweight.’’ Obesity: The Consequences and Naı¨ve Obesity Reviews 8:13–17. Assumptions Concerning the Causes of Krieger, Nancy. 1999. ‘‘Embodying Inequality: Physical Deviance.’’ Journal of Health and A Review of Concepts, Measures, and Social Behavior 21:75–87. Methods for Studying Health Consequen- Duster, Troy. 2006. ‘‘Comparative ces of Discrimination.’’ International Perspectives and Competing Explanations: Journal of Health Services 29:295–352. Taking on the Newly Reconfigured Lazarus, Richard S. and Susan Folkman. Reductionist Challenge to Sociology.’’ 1984. Stress, Appraisal, and Coping. New American Sociological Review 71:1–15. York: Springer. Ferraro, Kenneth F. and Jessica A. Kelley- LeBesco, Kathleen. 2004. Revolting Bodies? Moore. 2003. ‘‘Cumulative Disadvantage The Struggle to Redefine Fat Identity. and Health: Long-Term Consequences of Amherst, MA: University of Massachu- Obesity?’’ American Sociological Review setts Press. 68:707–29. Link, Bruce G. and Jo C. Phelan. 2001. Ferraro, Kenneth F. and Tetyana Pylypiv ‘‘Conceptualizing Stigma.’’ Annual Review Shippee. 2009. ‘‘Aging and Cumulative of Sociology 27:363–85. Inequality: How Does Inequality Get Long, J. Scott and Jeremy Freese. 2006. Under the Skin?’’ The Gerontologist Regression Models for Categorical 49:333–43. Dependent Variables Using Stata. Second Ferraro, Kenneth F. and Yan Yu. 1995. ‘‘Body Edition. College Station, TX: Stata Press. Weight and Self-Ratings of Health.’’ Maddox, George L., Kurt W. Back, and Journal of Health and Social Behavior Veronica R. Liederman. 1968. ‘‘Overweight 36:274–84. as Social Deviance and Disability.’’ Flegal, Katherine M., Margaret D. Carroll, Journal of Health and Social Behavior Cynthia L. Ogden, and Lester R. Curtain. 9:287–98. 2010. ‘‘Prevalence and Trends in Obesity McEwen, Bruce S. 1998. ‘‘Protecting and among U.S. Adults, 1999-2008.’’ Journal Damaging Effects of Stress Mediators.’’

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 96 Social Psychology Quarterly 74(1)

New England Journal of Medicine the Intersection of Gender, Race, and 338:171–79. Socioeconomic Status.’’ Journal of Mossakowski, Krysia N. 2003. ‘‘Coping with Gerontology: Social Sciences 62B:S415– Perceived Discrimination: Does Ethnic S423. Identity Protect Mental Health?’’ Journal Schnittker, Jason and Jane D. McLeod. 2005. of Health and Social Behavior 44:318–31. ‘‘The Social Psychology of Health Muennig, Peter. 2008. ‘‘The Body Politic: The Disparities.’’ Annual Review of Sociology Relationship between Stigma and Obesity- 31:75–103. Associated Disease.’’ BMC Public Health Schulze, Beate and Matthias C. Angermeyer. 8:128–38. 2003. ‘‘Subjective Experiences of Stigma: Muennig, Peter, Haomiao Jia, Rufina Lee, and A Focus Group Study of Schizophrenic Erica Lubetkin. 2008. ‘‘I Think Therefore I Patients, Their Relatives and Mental Am: Perceived Ideal Weight as Health Professionals.’’ Social Science & a Determinant of Health.’’ American Medicine 56:299–312. Journal of Public Health 98:501–506. Seeman, Melvin, Teresa Seeman, and Marnie National Heart, Lung, and Blood Institute. Sayles. 1985. ‘‘Social Networks and Health 1998. Clinical Guidelines on the Status: A Longitudinal Analysis.’’ Social Identification, Evaluation, and Treatment of Psychology Quarterly 48:237–48. Overweight and Obesity in Adults.National Sellers, Robert M., Cleopatra H. Caldwell, Institutes of Health Report No. 98–4083. Karen H. Schmeelk-Cone, and Marc A. Bethesda, MD: National Institutes of Zimmerman. 2003. ‘‘Racial Identity, Racial Health. Discrimination, Perceived Stress, and Neblett, Enrique W., J. Nicole Shelton, and Psychological Distress among African Robert M. Sellers. 2004. ‘‘The Role of American Young Adults.’’ Journal of Racial Identity in Managing Daily Racial Health and Social Behavior 43:302–17. Hassles.’’ Pp. 77–90 in Race and Identity: Stuber, Jennifer, Ilan Meyer, and Bruce Link. The Legacy of Kenneth Clark, edited by 2008. ‘‘Stigma, Prejudice, Discrimination, G. Philogene. Washington, D.C.: American and Health.’’ Social Science and Medicine Psychological Association Press. 67:351–57. Okosun, Ike S., Simon Choi, Tara Matamoros, Taylor, John and R. Jay Turner. 2002. and Alan Dever. 2001. ‘‘Obesity is ‘‘Perceived Discrimination, Social Stress, Associated with Reduced Self-Rated and Depression in the Transition to General Health Status: Evidence from Adulthood: Racial Contrasts.’’ Social a Representative Sample of White, Black, Psychology Quarterly 65:213–25. and Hispanic Americans.’’ Preventive Thoits, Peggy A. 1995. ‘‘Stress, Coping, and Medicine 32:429–36. Social Support Processes: Where Are We? Parker, Richard and Peter Aggleton. 2003. What Next?’’ Journal of Health and Social ‘‘HIV and AIDS-Related Stigma and Behavior Extra Issue: 53–79. Discrimination: A Conceptual Framework Wang, S. S., K. D. Brownell, and T. A. and Implications for Action.’’ Social Wadden. 2004. ‘‘The Influence of the Science & Medicine 57:13–24. Stigma of Obesity on Overweight Phelan, Jo C., Bruce G. Link, and John F. Individuals.’’ International Journal of Dovidio. 2008. ‘‘Stigma and Prejudice: One Obesity 28:1333–37. Animal or Two?’’ Social Science & Williams, David R. and Harold W. Neighbors. Medicine 67:358–67. 2001. ‘‘Racism, Discrimination, and Puhl, Rebecca and Kelly D. Brownell. 2001. Hypertension: Evidence and Needed ‘‘Bias, Discrimination, and Obesity.’’ Research.’’ Ethnicity and Disease 11: Obesity Research 9(12):788–805. 800–16. Puhl, Rebecca and Kelly D. Brownell. 2006. Williams, David R., Harold W. Neighbors, and ‘‘Confronting and Coping with Weight James S. Jackson. 2003. ‘‘Racial/Ethnic Stigma: An Investigation of Overweight Discrimination and Health: Findings from and Obese Individuals.’’ Obesity 14: Community Studies.’’ American Journal of 1802–15. Public Health 93:200–208. Schieman, Scott, Tetyana Pudrovska, and Winship, Christopher and Robert D. Mare. Rachel Eccles. 2007. ‘‘Perceptions of Body 1992. ‘‘Models for Sample Selection Bias.’’ Weight among Older Adults: Analyses of Annual Review of Sociology 18:327–50.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011 The Stigma of Obesity 97

BIOS Kenneth F. Ferraro is Distinguished Professor of Sociology and director of the Markus H. Schafer is currently Center on Aging and the Life Course at a PhD candidate at Purdue University; Purdue University. Recent publications later this year he will be joining appear in the American Journal of the University of Toronto as an assistant Sociology, American Journal of Public professor of sociology. His research Health, Social Science and Medicine,and interests include subjective aspects The Gerontologist. With interests in how of aging, physical and social consequen- stratification processes unfold over the ces of obesity, and the connections life course, he has developed cumulative between childhood adversity and adult inequality theory and is now directing health. research projects to test elements of it.

Downloaded from spq.sagepub.com at ASA - American Sociological Association on March 4, 2011