Read the sociology article and write response
The Unequal Weight of Discrimination: Gender, Body Size, and Income Inequality Author(syf Katherine Mason Reviewed work(syf :
Source: Social Problems, Vol. 59, No. 3 (August 2012yf S S 5 Published by: University of California Press on behalf of the Society for the Study of Social Problems Stable URL: http://www.jstor.org/stable/10.1525/sp.2012.59.3.411 .
Accessed: 10/11/2012 11:49 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp .
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
. University of California Press and Society for the Study of Social Problems are collaborating with JSTOR todigitize, preserve and extend access to Social Problems. http://www.jstor.org The Unequal Weight of Discrimination:
Gender, Body Size, and Income Inequality Katherine Mason, University of California, Berkeley At present, most work examining the well-documented relationship between social inequality and body size treats fatness as an effect, caused either by some factor that determines weight and social class simultaneously, or by social class itself. However, the relationship between weight and social inequality is more complex than these explan- ations suggest. Recent studies by John Cawley (2004) and Charles Baum and William Ford (2004) suggest that fatness is often a contributor to inequality, not merely an effect.
This article examines the causes of income inequalities between obese and nonobese workers, focusing on how gender interacts with body size to determine the size and duration of those inequalities. Drawing on data from the 1997–2008 National Longitudinal Survey of Youth (NLSY97), I introduce a positive test for discrimination, which provides a methodological advantage over previous research in this area. I then pose two questions: first, is anti-obesity discrimination to blame for income inequalities between obese and nonobese workers? Second, do women and men’s experiences of those inequalities differ? The results indicate that very obese men do face one form of discrimination— statistical discrimination—but that they can overcome initial disadvantages with time. In contrast, obese women’sin- come disadvantages persist over time, suggesting the presence of prejudicial discrimination. In combination with pre- vious studies illustrating how fat women are disadvantaged in educational attainment and marriage outcomes—two important means of accessing economic resources—this research shows one mechanism by which weight, particularly in combination with gender, is a major vector of U.S. inequality. Keywords: income inequality; overweight; gender; statistical discrimination; prejudicial discrimination. Americans are getting fatter. 1This oft-repeated claim, usually presented as a simple state- ment of fact, tends to evoke a particular range of anxieties, meanings, and moral judgments, near- ly all of which assert that it is bad to be fat. To public health researchers, the statistic that more than half of the adult population of the United States has been classified as“overweight”or“obese” since the mid-1980s is cause for alarm. These researchers predict that current generations will live shorter lives than their parents and that those lives will be marked by greater rates of ill health and disability. Accordingly, public policy analysts worry that such predicted health consequences will lead to increased health care costs. Former U.S. Surgeon General Richard Carmona exclaimed that“[o]besity is a terror within; it’s destroying our society from within and unless we do some- thing about it, the magnitude of the dilemma will dwarf 9/11 or any other terrorist event that you can point out”(Pace 2006). In contrast, many others, particularly those outside of the The author is grateful to Michael Hout, Samuel Lucas, Kristin Luker, Raka Ray, and Barrie Thorne for their thoughtful suggestions on this project. She also thanks the faculty and graduate student members of the Berkeley Sociology Gender Workshop, whose close reading and commentary on this article helped in specifying its theoretical concepts. Lastly, special thanks to theSocial Problemseditors and anonymous reviewers for challenging the author to make this project as time relevant as possible. Direct correspondence to: Katherine Mason, Department of Sociology, University of California, Berkeley, 410 Barrows Hall, Berkeley, CA, 94720–1980. E-mail: [email protected]. 1. Throughout this article, I use the term“fat”not in the pejorative sense, but instead as it has been reclaimed by fat activists (similar to the use of the word“queer”by LGBT groups). To this end,“fat”points to the aesthetic and normative oth- erness that large-bodied people experience, rather than to the medical criteria that classify them as“overweight.”I use the terms“overweight”and“obese”when I am referring to body mass index designations or when they are the wording preferred by an author I am discussing. Social Problems, Vol. 59, Issue 3, pp. 411–435, ISSN 0037-7791, electronic ISSN 1533-8533. © 2012 by Society for the Study of Social Problems, Inc. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content through the University of California Press’s Rights and Permissions website at www.ucpressjournals.com/reprintinfo/asp.
DOI: 10.1525/sp.2012.59.3.411. United States, view fatness as a fitting metaphor for America’s culture of expansionism and overconsumption (Saguy 2008). In this article, I do not examine the claim that being fat is bad for one’s health, nor do I document the range of discourses and meanings surrounding fatness in the United States (for coverage of such discourses, see Boero 2007 and Saguy and Riley 2005).
I instead focus on how, as carriers of the so-called“obesity epidemic,”fat people embody socially stigmatized characteristics (Carr and Friedman 2005; Puhl and Heuer 2010) and thereby face a range of social disadvantages including—but not limited to—income discrimination. This article highlights the shortcomings of how weight-based discrimination has been measured in past research, and proposes a method for positively establishing the presence of discrimination. I then use this method to test past research’s contention that women are more severely disadvantaged by weight-based inequality than are men, while updating the literature with recent data from the 1990s and late 2000s.
In taking this approach, my work joins a small but growing body of research that examines the income inequalities between fat and nonfat people. These studies step outside of typical fram- ings of fatness and fat people as problems for society, instead analyzing how existing social arrangements are problematic for fat people. Namely, as anti-fat rhetoric (such as the perspectives described above) characterizes fatness as a stigmatized property of individuals, people with that characteristic may face discrimination and inequality due to their body size.
As a note about terminology, the convention in studies such as this is to discuss body size in terms of the four categories of the body mass index (BMI): underweight, normal weight, over- weight, and obese. Some studies (e.g., Conley and Glauber 2007) also use the term“healthy weight”in place of“normal weight.”In this article, I will use the BMI categories when I am refer- ring to body mass specifically as it is measured by height and weight: this is the metric available to me in the data I analyze. When speaking about weight-based inequality in general, however, Iwillusetheterms“fat”and“nonfat.”This language choice reflects the fact that there is consid- erable disagreement both within and outside of medical circles about the exact relationship between body mass and health. At the start of the twenty-first century, people who fall within the “normal”BMI range are no longer the norm—as of 2006, adult men in the United States had an average weight of 195 pounds and an average BMI of 28.8 (well into the overweight range), while women averaged weights of 165 pounds and BMIs of 28.3 (McDowell et al. 2008). Furthermore, using the term“healthy weight”to describe people with BMIs in the 18.5–25 (“normal”)range may also be inaccurate. Several studies, most notably Katherine Flegal and colleagues’(2005) article,“Excess Deaths Associated with Underweight, Overweight, and Obesity,”have shown that it is in fact people with BMIs in the“overweight”range, not the“normal”range, who have the lowest rates of mortality and morbidity in the United States. For this reason, when not referencing BMI in particular, I use descriptive terms—“fat”and“nonfat”—in place of the normative language of“normal”or“healthy ”weight.
Previous research in this area has investigated fat people’s personal experiences of discrimi- nation and its relationship to self-esteem (Carr and Friedman 2005), the legal and ethical guide- lines for dealing with anti-fat employment discrimination (Kristen 2002; Roehling 2002), and the correlation between fatness and lowered wages and rates of employment, often mediated by gen- der and/or occupation (Averett and Korenman 1996; Cawley 2004; Cawley and Danziger 2005; Gortmaker et al. 1993; Hamermesh and Biddle 1994; and Haskins and Ransford 1999).
While many of these studies provide quantifiable evidence of the employment inequality faced by fat people, they are nonetheless limited in what they can tell us about the nature of that inequality. Researchers may claim that the income differences between fat and nonfat people are due to discrimination, but they face countervailing arguments that fat people are lazy, undisci- plined, and thus, less likely to succeed in well-paying jobs. While such arguments may strike some as being premised on inaccurate stereotypes, they nevertheless point to methodological weak- nesses in the aforementioned studies. Several researchers have tried to control for confounding traits by including subjects’intelligence test scores or educational achievements in their analyses (e.g., Averett and Korenman 1996; Gortmaker et al. 1993), but so long as they determine 412MASON discrimination via residual explanations—that is, by controlling for a number of variables that might explain differences in income and then taking any remaining income differences between groups as evidence of discrimination—they are always open to the critique that those differences are caused not by anti-fat discrimination, but by some other explanatory factor left out of the analysis.
The purposes of this study are twofold. First, I update the research on weight-based income inequality, most of which was published in the mid-1990s and based on data from the 1980s (such as Steven Gortmaker and colleagues’[1993] study using the 1979–1988 National Longitudinal Survey of Youth, or NLSY). My data, a comparable sample of young people born approximately two decades after the subjects of the previous wave of research (the 1997–2008 National Longitu- dinal Survey of Youth, hereafter NLSY97), allow me to revisit the hypotheses that earlier research- ers tested while paying heed to changing social norms across the intervening years (most notably, trends in body size, marriage, and educational attainment). The second purpose of this study is to elaborate apositivetest for a specific type of discrimination: statistical discrimination. While statisti- cal discrimination is not the only type of discrimination that exists, this test (first described by Joseph Altonji and Rebecca Blank [1999] for studying racial/ethnic and gender inequalities) makes it possible to positively establish the presence of discrimination, simultaneously providing a more nuanced view of the different forms and degrees of discrimination that may be operating.
Using the concepts ofmeritocratic discrimination, statistical discrimination,andprejudicial discrimination as tools for investigating the nature and severity of income disparities among obese and nonobese workers, this study provides evidence that (1) weight-based income discrimination against obese people exists, (2) its effects are especially pronounced among those who are very obese, and (3) it has more severe consequences for women than for men.
The remainder of this article is divided up as follows. First, I discuss previous work showing that the fat body is often subjected to workplace discrimination. Several of these studies have highlighted how women, especially, are the victims of anti-fat prejudice (even when anti-fat dis- courses do not explicitly target them). In a discussion of the particular article on which my own work is premised, Gortmaker and colleagues’(1993)“Social and Economic Consequences of Overweight in Adolescence and Young Adulthood,”and the accompanying response from the editors at theNew England Journal of Medicine, I show why a residual explanation of discrimination is insufficient, and why a positive test for discrimination is needed. Further, I elaborate a spectrum of discrimination against which to compare my findings, based on varying degrees of employer rationality and stereotyping. Subsequently, I lay out the four hypotheses to be tested, describe the methods and data used in this study, and review the findings. Finally, in the discussion, I explain the significance of these results and suggest future directions for research. Fatness and Income Inequality: A Review What Is the Relationship between Fatness and Socioeconomic Inequality?
The existing studies on anti-fat discrimination in employment provide a helpful framework for thinking about the problem overall. First, by discussing the correlation between obesity and experiencing inequality in the form of lowered wages, this article asserts that fatness is often a causeof diminished income and life chances, not solely a consequence.
Although the association between lower socioeconomic status (SES) and fatness in the United States has been well documented, the direction of causation in that correlation remains hotly debated. Unlike characteristics such as race/ethnicity and sex, both of which are popularly seen as stable characteristics of an individual, both class and body size are viewed as being somewhat vari- able at the individual level (reflected in the fields of study dealing with, respectively, social mobility and weight management). Thus, whereas a correlation between, say, sex and income could be easily read as showing the effect of one’s sex (the unchanging, causal variable) on one’sincome, causality is less apparent in the relationship between body size and class. Cawley (2004) outlined The Unequal Weight of Discrimination413 and tested three major hypotheses surrounding the correlation between obesity and low wages:
(1) obesity causes lower wages (whether due to decreased productivity or to discrimination); (2) low wages cause obesity (because of the availability of cheap, bad-quality, fattening foods); and (3) some other factor (such as, for example, inherited social class) causes both obesity and low- er wages. Cawley found no evidence to support hypothesis 2, but he did find evidence to support hypotheses 1 (particularly for women) and 3 (in the case of certain racial/ethnic minority groups).
The connections between obesity, SES, stress, health, and nutrition (to name just a few) are, no doubt, complex and often multidirectional. What Cawley found is that the relationship between obesity and income cannot be explained completely by background factors, and the direction of causation in that relationship is from obesity to income, not the reverse.
If fatness (at least in the United States) contributes to income stratification, what does that phenomenon look like, and how does it play out? Furthermore, how does gender mediate the re- lationship between weight and income? In the following section, I review the three main meth- odological avenues through which researchers have approached these questions: (1) laboratory studies of discriminatory hiring and employment practices toward hypothetical employees; (2) self-reported experiences of stigma and discrimination; and (3) residual explanations of dis- crimination in large-scale multivariate statistical analyses. Thereafter, I describe how my chosen approach will build on existing studies’findings while adding a more robust measure of employ- ment discrimination into the mix. Laboratory Studies Cawley’s research, which looks at discrimination in real-life statistics, is rare among studies in this field; more common are experimental studies of discrimination such as those described by Rebecca Puhl and Kelly Brownell (2001):
Studies on employment have shown hiring prejudice in laboratory studies. Subjects report being less in- clined to hire an overweight person than a thin person, even with identical qualifications. Individuals make negative inferences about obese persons in the workplace, feeling that such people are lazy, lack self-discipline, and are less competent. One might expect these attributions to affect wages, promotions, and disciplinary actions, and such seems to be the case (p. 800). While such studies provide a helpful insight into the possible workings of weight-based employ- ment inequality, Puhl and Brownell note that they are limited by their artificiality—this type of research has tended to be based on the responses of a convenient study population, college stu- dents, making hiring judgments about hypothetical employees, not on the actual practices of sea- soned employers.
In a similarly controlled laboratory experiment, Matthew Mulford and associates (1998) found that people tended to cooperate better and more frequently with coworkers they perceived to be attractive, a judgment that, at least in the U.S. context, is likely to include the evaluation of people’s bodies. Likewise, Daniel Hamermesh and Jeff Biddle (1994) discovered that there were both earnings penalties for unattractiveness and earnings premiums for exceptional physical beauty, regardless of one’s occupational category or gender. They did find other effects by gender, though, showing that women who were considered unattractive participated in the labor force at lower rates and married men with less human capital. In a separate study, these authors employed a four- person panel of judges to rate the matriculation photographs of lawyers for physical beauty and found that those deemed more attractive were more often selected into private practice (i.e., higher-paying) law jobs than those who were less attractive (Biddle and Hamermesh 1998). Self-Reports of Weight-Based Stigma and Discrimination In addition to the aforementioned laboratory experiments that purport to measure average people’s propensity to discriminate on the basis of weight or attractiveness, another type of 414MASON research that has contributed to scholarly understandings of weight-based discrimination involves the use of self-reported experiences of stigmatization and inequality. Puhl, Tatiana Andreyeva, and Brownell (2008) looked at data from the 1995–96 National Study of Midlife Development in the United States (MIDUS), which asks respondents to report on any discrimination they have en- countered in a range of settings. The authors found significant gender differences in the preva- lence of reports of height/weight discrimination (4.9 percent of men reported being affected by height/weight discrimination, while more than double that number of women—10.3 percent— did so). They also found important gender differences in how educational background and age af- fected respondents’likelihood of reporting height/weight discrimination, and noted that height/ weight discrimination was comparable to racial/ethnic discrimination in its prevalence. Deborah Carr and Michael Friedman (2005), who also analyzed the MIDUS 1995–96 data, add that better-educated respondents and those in higher-status occupations (professional and managerial positions) were more likely to report being victims of discrimination.
Markus Schafer and Kenneth Ferraro (2011), who used the MIDUS 1995–96 data as well, ar- gue that whether or not the experience of discrimination that respondents report is objectively “real,”itseffectscertainly are. Drawing on the sociological concept of“stigma,”they suggest that respondents’understandings of their own bodies are mediated through social perceptions, such that the feeling of stigmatization may actually have a negative effect on respondents’physical well-being. Puhl and Chelsea Heuer (2010) agree, writing that the“stigmatization of obese indi- viduals poses serious risks to their psychological and physical health, generates health disparities, and interferes with implementation of effective obesity prevention efforts”(p. 1019).
Studies based on self-reported experiences of weight-based discrimination provide an impor- tant perspective on how negative societal attitudes toward fat bodies may become the basis for stigmatization and a wealth of negative consequences that follow. In such research (Puhl and Heuer 2010; Schafer and Ferraro 2011), theobjectiveexperience of discrimination (as measured by, for example, material inequities) is less important than the respondents’subjectivefeeling of stigma. However, this characteristic also represents a limitation in what self-reported weight discrimination studies can tell us. In the third section of this literature, then, I turn to large-scale surveys that study the income disparities between obese and nonobese workers. Residual Explanations of the Relationship between Fatness and Income Inequality Baum and Ford (2004) draw on nearly two decades’worth (1981–1998) of longitudinal data from the original National Longitudinal Survey of Youth (NLSY) to look at earnings inequalities due to body size. They find that there is a robust wage penalty for being obese (significant in a va- riety of statistical models that they test) and, in keeping with previous findings about the effects of gender, that women consistently suffer more from this penalty than do men (p. 897). Although the authors suggest that at least some of this inequality is due to employer discrimination, they add that they did not test for it. At most, they make a convincing case for discrimination using a residual explanation (that is, attributing to discrimination residual differences between groups that remain after controlling for other likely causes in a regression model).
Susan Averett and Sanders Korenman (1996), also using data from the NLSY, found that obese women tended to reside in lower socioeconomic brackets than nonobese women (which was not the case for obese men), and that this phenomenon was due to a combination of out- comes of obesity: lowered marriage prospects (by which obese women had less chance of marry- ing and, if married, less chance of being married to a man with high earnings), as well as direct employment discrimination.
Gortmaker and colleagues (1993) conducted one of the most comprehensive statistical stud- ies of body size and income inequality. They suggest a starting model for testing income inequal- ities in overweight and nonoverweight individuals using the NLSY, an earlier wave of the data set (NLSY97) I analyze in this project. Examining variables from the 1988 data, Gortmaker and asso- ciates found persistent, significant effects of being“overweight”in several areas of adult life: for The Unequal Weight of Discrimination415 men, lowered incomes; for women, not only lowered incomes but also lower rates of marriage and lower educational attainment. While this study did not prove conclusively that discrimination was the cause of these differences, it did rule out several competing theories. First, the authors found no support for the hypothesis that differences in economic outcomes between overweight and nonoverweight people could be explained by their class origins (a hypothesis that Cawley [2004] also examined and rejected). In other words, even though people brought up in poorer households and neighborhoods were more likely to be overweightandto work in low-income jobs, accounting for class of origin could not explain away the association between overweight and lower incomes. A second hypothesis this article tested was that overweight people might be less healthy than nonoverweight people, and that chronic ill health—not discrimination—might be the cause of the lower-paying jobs held by overweight respondents. This hypothesis, too, was unsupported by the evidence. Likewise, overweight people did not appear to differ markedly from nonoverweight people in their self-esteem, and so their disparate financial outcomes could not be explained by personal characteristics like confidence and self-esteem. From these findings, the au- thors concluded that the most likely causes of overweight people’s lower incomes were stigma and discrimination.
Although Gortmaker and his colleagues rigorously ruled out several competing theories be- fore they arrived at this conclusion, the response from the editors in the same issue of the New England Journal of Medicineshows why even a very well done residual account of discrimina- tion may not be enough to convince skeptics. Albert Stunkard and Thorkild Sorensen (1993), the editors who wrote a comment on the piece by Gortmaker and associates, suggested that further research should take into account the possible influences of genetics on both SES and weight. On the face of it, this suggestion seems reasonable; previous research has found that children may inherit their biological parents’SES even when those parents do not raise them, perhaps through mechanisms like IQ. Stunkard and Sorensen noted that there is a good chance that children’s weight and body types may also contain a genetic component inherited from their parents. What makes this editorial troublesome, however, is that genetics were offered as a possible explanation of not just SES or fatness, but theconnectionbetween the two. In other words, they seem to be sug- gesting that genetic determinants of SES (such as intelligence) may be tied to genetic determinants of weight, thus implying that fat people are less intelligent than nonfat people. That the editors could make this connection in spite of Gortmaker and his colleagues having controlled for intelli- gence (and finding that income differences between overweight and nonoverweight people persisted) shows that still more rigorous testing for discrimination may be needed.
Thus, the three aforementioned branches of research on weight and income inequality— laboratory experiments measuring test subjects’propensity to discriminate against hypothetical fat employees, subjective self-reports of discrimination and stigmatization on the basis of one’s weight, and multivariate analyses using residual explanations of discrimination—all provide evi- dence that strongly suggests the existence of anti-fat discrimination. However, none employs the methodological tools to positively establish and quantify the existence of such discrimination in real-world employment settings. It is this gap that this article seeks to address. When Do Income Differences Constitute Unlawful Forms of Discrimination?
While Stunkard and Sorensen’s (1993) implication that genetics and intelligence may partially explain fat people’s lowered SES is problematic, it also points to the notion that not all differences in income result from unlawful discrimination. Most employers can be said to discriminate in their employment practices, in the sense of the word meaning to distinguish (and choose) be- tween better and worse options. What is at stake, then, is not whether employers discriminate, but how rational and lawful we perceive their discriminatory criteria to be. I argue that it is pro- ductive to think of these criteria as falling on a spectrum ranging from rational and legitimate on one side to irrational and illegitimate on the other. These assessments of each type of discrimina- tion’s legitimacy are not absolute, but are contingent on capitalist market logic (which supports 416MASON merit-based hiring and wages over, for example, need-based pay) as well as on the particular requirements of a given job.
At one end of the spectrum, income differences may be due to differences in experience, skill, or effort. Here, the cause is what I am callingmeritocratic discrimination. A classic example of this form of inequality can be found in Emile Durkheim’s ([1933] 1997) discussion of the division of labor, where he writes,“Labour only divides up spontaneously if society is constituted in such a way thatsocial inequalities express precisely natural inequalities”(p. 313; emphasis added). In a capital- ist society, we would not consider socioeconomic differences due to productivity and skill to be illegitimate in and of themselves, even if the conscientious social scientist should be concerned with whether opportunities for people to gain the necessary experience, to identify and develop their skills, and to have their efforts recognized are distributed equitably. In the ideal-typical form of meritocratic discrimination, wage and employment differences among individual workers would neatly reflect their productivity and the demand for their labor, as given by theories about human capital and a well-functioning free market (e.g., Becker 1975; Medoff and Abraham 1981).
A second form of discrimination, which falls in the middle of this spectrum, is Lester Thurow’s (1975) notion ofstatistical discrimination. In the case of statistical discrimination, employers attempt to make rational hiring and pay decisions based on spotty knowledge of an individual worker’smerit.People who belong to groups believed to lack favorable work characteristics are paid less or“are not hired because of the objective characteristics of the group to which they belong, although they, themselves, are satisfactory”(p. 172). In this form of discrimination, cer- tain groups may have a loweraveragechance of possessing some trait that an employer is looking for (e.g., one might expect that people with lower levels of education would have, on average, lower levels of reading comprehension and writing skills), but that expectation does not hold true for every member of the group. Such outstanding individuals are said to suffer discrimination as a consequence of belonging to an“objectively”less desirable group. How can statistically undesir- able individuals overcome such profiling? Thurow (1975) states: The acceptable workers buy their way out of the group to which they belong. Individuals know whether they do or do not have the desired characteristic. If they do, they can overcome their group’s character- istics by offering to work for a short period of time for a wage lower than others who are believed to have the right set of personal characteristics. Then once on the job, where they can demonstrate that they have the right characteristics, their wages will rise to the level of others with the right characteristics regardless of the groups to which they belong (p. 173). Thurow’s argument rests on the rather tenuous premise that, in many cases,“objective”statistical information about group characteristics is obtainable. It seems likely, however, that employers will inevitably substitute well-researched information with“common knowledge.”In the case of fat jobseekers, many employers may imagine them to embody traits like“laziness, lack of disci- pline, unwillingness to conform, and absence of all those‘managerial’abilities that, according to the dominant ideology, confer upward mobility”(Bordo [1993] 2003:195), and may also view them as“lacking self-discipline, having low supervisory potential, and having poor personal hy- giene and professional appearance”(Puhl and Brownell 2001:789).
For Thurow, the problem of statistical discrimination results from employers’rational at- tempts to make employment decisions efficiently: they supplement their limited observable infor- mation about each applicant with statistical knowledge that they believe accurately describes a group but may be inaccurate for any individual member of the group. Statistical discrimination does not describe a situation in which members of one group systematically exploit or deny em- ployment to members of another group. Rather, it assumes that employers are motivated by the desire to find workers who provide good value while minimizing the costs of searching for and identifying qualified job candidates. When potential employers are unable to ascertain whether an applicant has certain desired worker characteristics like dedication or intelligence (an issue that arises particularly in the case of young or inexperienced jobseekers), they fall back on observable characteristics as crude proxies. The Unequal Weight of Discrimination417 Finally,prejudicial discriminationfalls at the opposite end of the spectrum from meritocratic discrimination. It entails employment decisions based on characteristics that are unrelated to worker skill or productivity, and it is the form of discrimination most commonly named as such in popular and scholarly discourses. When prejudicial discrimination comes into play, it makes employers blind (or indifferent) to the qualifications of people from stereotyped groups, with the result that such job applicants will be (1) not hired or (2) placed in jobs for which they are over- qualified or ill suited. As such, prejudicial discrimination represents the most blatant and irrational contributor to income inequalities among individuals in a capitalist economy.
Thinking about wage disparities in terms of a spectrum—ranging from meritocratic to statis- tical to prejudicial discrimination—is more helpful for developing a nuanced vision of how discrimination works than simply looking for the presence or absence of discrimination. Specifi- cally, while these forms of discrimination may differ in their consequences for both employers and employees, all arise from employers’varying ability (or willingness) to recognize employee pro- ductivity and to match them to appropriate jobs and wages. Furthermore, the three forms of dis- crimination described above are given as ideal types, but in practice they are not always clear-cut or mutually exclusive. Thurow (1975:175), for example, notes that if those who seek to benefit from another group’s exclusion from a workplace or occupation can manage to reduce the qual- ifications of enough members of that group (such as by defunding schools in regions where mem- bers of that group are concentrated), employers may inadvertently help out the exploiting group by starting to practice statistical discrimination against the exploited group. Such a situation would fall somewhere between statistical and prejudicial discrimination on the spectrum. On the other side of things, if, for example, an employer advertised a job for which punctuality was important, s/he might ask job applicants’past employers to comment on the individual’s history of coming to work on time, and make a hiring decision on that basis. This would be a simple case of meritocratic discrimination. If, however, the job opening was for an entry-level position and no direct evidence about applicants’past punctuality was available, the employer might turn to whatever proxies s/he could find: for example, whether the employee was single or married, with or without chil- dren. Although regulations such as the U.S. federal Equal Employment Opportunity laws draw a sharp line between the first case (meritocratic discrimination) and the second (statistical discrimi- nation, which violates the law), understanding these cases in terms of a spectrum of discrimina- tion highlights the ways in which efforts at finding and rewarding the best workers can go wrong when employers have limited knowledge.
While one point of this spectrum model is to emphasize the continuities between different forms of discrimination, the three particular forms I have specified are helpful tools for determining the severity of inequality: not all discrimination iscreated equal. One difference between statistical and prejudicial discrimination is in the scope of their effects. Whereas statistical discrimination indi- cates unfair treatment of individuals (but the majority of group members for whom the stereotype is true are being treated fairly according to their value as workers), prejudicial discrimination has the potential to unfairly disadvantage all members ofa given group. The othermain difference between statistical and prejudicial discrimination becomes evident over time. Under statistical discrimination, wage deficits for low-status people should catch upto those of other comparably qualified employees as employers gain better information about them,even though their lifetime earnings will be less than those of their colleagues. On the other hand, under prejudicial discrimination, low status wages will not onlyfailto catch up to high status wages, but the effects of such discrimination will greatly accumulate over time and widen the gap. Neither ofthese two forms of discrimination is legally de- fensible, nor is it likely that workers’losses from either form will be fully recouped over time, but the magnitude of those losses differs substantially between the two types.
Developing a Positive Test for Statistical Discrimination Joseph Altonji and Rebecca Blank (1999) suggest a quantitative method of testing for the presence of statistical discrimination, which they recommend for studies of race and gender 418MASON inequality. Many studies of discrimination, including the previously mentioned work by Gortmaker and associates (1993), look for discrimination in the residual differences in income or employment that remain after controlling for all the relevant variables available. As we have seen, the problem with this type of research is that it can never fully rule out the possibility that income differences are simply the result of some legitimate difference in workers’skill or productivity that is visible to employers, but unobserved by the researcher (in other words, meritocratic discrimina- tion). In contrast, Altonji and Blank’s model is appealing because it finds a way topositivelytest for statistical discrimination. Using an interaction term that multiplies one’s years of work experience by the variable believed to be a basis for discrimination, Altonji and Blank test whether the nega- tive effects on income due to the possession of that characteristic decrease as experience grows. If those effects lessen over time, statistical discrimination is indicated: employers gradually replace their statistical knowledge about the group to which the worker belongs with knowledge about the individual worker, and the worker is compensated accordingly. If, instead, earnings inequal- ities persist despite work experience, they are more likely due to prejudicial discrimination. Hypotheses In this article, then, I set out to determine the nature and magnitude of the income disparities faced by obese employees using Altonji and Blank’s (1999) positive test of statistical discrimina- tion. Through multivariate ordinary least squares (OLS) regression, I test the following hypotheses:
H0: Controlling for a variety of background factors, particularly including education level and intelligence test results, will cause the relationship between obesity and reduced income to disappear. This finding would indicate the presence ofmeritocratic discrimination.
H 1: Disparities in income between obese and nonobese individuals will persist after controlling for a va- riety of common causes of income inequality. Furthermore, there will be a significant interaction effect between work experience and obesity, such that the income gap between obese and nonobese workers will decrease as work experience increases. This finding would indicate thatstatistical discriminationis at work.
H 2: Income disparities between obese and nonobese workers will exist, but there will be no significant interaction between obesity and work experience. This outcome would suggest that the obese worker’s disadvantage in any given year neither increases nor lessens with time. There would, nonetheless, be an accumulation of lost wages from year to year, and it would be strongly suggestive ofprejudicial discrimination.
Finally, most existing studies in this field have tested for—and confirmed—gender differences in the degree to which fat women and men face employment inequality relative to their thinner colleagues. These findings fit with gender scholarship arguing that women and men often face dif- ferent types and degrees of social scrutiny in regard to their bodies. Gendered ideals of masculine and feminine characteristics map onto bodies, creating expectations that, for example, men will be heavier and taller, and will require more space than women (Bordo [1993] 2003); men eat heart- ier food in greater portions while“a girl’s accession to womanhood is marked by doing without” (Bourdieu [1979] 1984:195); men have the right to look at and judge women’s bodies (while women do not have parallel visual access to men’s bodies) (Berger [1972] 1977; Mulvey 1975); and women, freighted with historical and ideological ties to the body and nature (often evoked through associating womanhood with childbirth and maternal“instincts”or by contrasting men’s supposedly innate intellectuality with feminine“nature”), are thus more likely than men to be evaluated according to their physical characteristics (Butler [1990] 2006; Ortner 1974; Wolf [1991] 2002). In light of these patterns, an additional hypothesis is needed:
H3: Women and men may have differences in the quality and/or quantity of discrimination they face, with anti-obesity discrimination tending to be more damaging to women than to men.The Unequal Weight of Discrimination419 Data Source and Variables Several studies providing residual explanations of discrimination (Averett and Korenman 1996; Gortmaker et al. 1993) have used the National Longitudinal Survey of Youth (NLSY), a sur- vey that has been administered annually by the U.S. Bureau of Labor Statistics (BLS) to a nation- ally representative sample of 12,686 people since 1979. The large, representative nature of the sample, its longitudinal design, and its wide variety of attitudinal, background, and lifestyle vari- ables make it ideal for this type of study. However, these authors’findings on weight-based in- come inequalities, using data from the 1979–88 rounds of the NLSY are now more than two decades old. Given the explosion of interest in (and concern about) fatness in the American pop- ulation in the intervening years, examination of a more current set of data is long overdue. Thus, I will base my study on an examination of the 1997–2008 NLSY97 data. In 1997, the BLS initiated a second wave of the NLSY (NLSY97), sampling a new group of 8,984 youth between the ages of 12 and 17. In addition to baseline characteristics derived from 1997 variables, including respond- ents’primary racial/ethnic identification, sex, logged household income, and parents’educational attainment, I also used items from the 1999, 2001, and 2008 rounds of the NLSY97 to construct my variables of interest. Since not all original respondents participated in these follow-up years, I obtained sample weights from NLSY97 for only those individuals who had data for 1997, 1999, 2001, and 2008. Of those people who answered all of the variables of interest, 1,196 were women and 1,231 were men. In general, questions from later waves of the study had higher rates of missing data than those drawn from earlier waves, likely due to subjects moving or dropping out of the study, and it is largely these later questions that are responsible for the sample size in the current study. Summary statistics for all variables listed below can be found in Table 1.
The research by Gortmaker and associates (1993) suggests a number of helpful control vari- ables for the analysis. These control variables include: race/ethnicity, gender, socioeconomic back- ground (measured by respondents’1997 household income and the number of years of schooling completed by each parent), and several personal characteristics (health limitations on work ability, intellectual ability as measured by the Armed Services Vocational Aptitude Battery, or ASVAB, and respondent’s educational attainment).
To these controls I added a variable for whether the respondent was obese or not in 2001 (lagged seven years behind the year in which income was measured), a variable for whether the respondent was a parent in 2008, a measure of work experience, and a variable for occupational status. The operationalization of selected variables will be addressed below. Finally, respondents’ logged net income in 2008 served as the dependent variable.
Variables of Interest Income.Income, the dependent variable, was measured in logged 2008 dollars 2for the past year. Incomes from personal farm and business ventures were excluded since one would not expect the self-employed to discriminate against themselves. Because the respondent’sprevious year’s income was the dependent variable, I further restricted this analysis to people who were in the workforce in the preceding year (defined by having received any income from employment). 3 Obesity.To measure whether respondents were obese or nonobese, I used questions asking for self-reported weight and height, in pounds and inches. I then calculated individuals’ body mass index (BMI) scores as a ratio of weight and height using the following equation:
2. Logging income helps correct for skewness in the variable.
3. Another control variable in my regression model asks whether someone has any substantial health issues that affect one’s ability to work. Many (Gortmaker et al. 1993) have noted that fat individuals might be more likely to suffer from such health issues. For this reason, it is important to check whether limiting the regression to people in the workforce is systematically excluding fat individuals on the basis of health limitations. 420MASON BMI=(kg/m 2).4Although BMI does not take into account the difference between muscle mass and fat, studies have shown that it correlates well with other, more involved measures of body fat composition for the majority of people (Gortmaker et al. 1993). Similarly, while BMI is an inexact measure of fatness for any given individual, it is a serviceable proxy at the aggregate level.
BMI also has the advantage of being composed of two variables that are easy to measure: height Table 1Descriptive Statistics for U.S. Women and Men, 1997–2008 Women Men Categorical VariablesNPercentNPercent 1,196 100 1,231 100 Primary race/ethnicity (1997) Black 248 20.74 173 14.05 White 716 59.87 825 67.02 Hispanic 179 14.87 200 16.25 Other 53 4.43 33 2.68 BMI category (2001) Underweight (BMI < 18.5) 66 5.52 35 2.84 Normal weight (18.5 < BMI < 25) 756 63.21 700 56.86 Overweight (25 < BMI < 30) 233 19.48 325 26.40 Obese (BMI > 30) 141 11.79 171 13.89 Obese I (30 < BMI < 35) 91 7.61 122 9.91 Obese II/III (BMI > 35) 50 4.18 49 3.98 College diploma or higher? (2008) Yes 478 39.97 329 26.73 No 718 60.03 902 73.27 Any health limitation on work? (2008) Yes 544.52 493.98 No 1,142 95.48 1,182 96.02 Any children? (2008) Yes 482 40.30 403 32.74 No 714 59.70 828 67.26 Upper white-collar occupation? (2008) Yes 573 47.91 403 32.74 No 623 52.09 828 67.26 Continuous VariablesNMean(SD) NMean(SD) Logged household income per capita (1997)1,196 10.59 (.93) 1,231 10.60 (.94) Mother’s years of education completed (1997)1,196 13.05 (2.78) 1,231 13.05 (2.74) Father’s years of education completed (1997)1,196 13.03 (3.00) 1,231 12.97 (3.08) Armed services vocational aptitude battery test (1999)1,196 54.50 (27.13) 1,231 52.58 (28.85) BMI in kg/m 2(2001) 1,196 24.08 (5.08) 1,231 24.87 (4.73) Age in years (2008) 1,196 25.88 (1.42) 1,231 25.79 (1.43) Total weeks worked (2008) 1,196 349.08 (112.83) 1,231 334.02 (116.24) Logged respondent income (2008) 1,196 9.83 (.96) 1,231 10.12 (.83) Source: National Longitudinal Survey of Youth 97 (1997, 1999, 2001, and 2008 cases).
Note: Variable names are accompanied by the survey year from which the data were drawn.
4. Conversion from standard measures to metric uses the formula:kg/m 2=lbs/in 2*703. The Unequal Weight of Discrimination421 and weight. The second National Health and Nutrition Examination Survey (NHANES II) found that self-reported and measured height and weight among subjects in their early twenties differed significantly for only 3 percent of women and 1 percent of men, leading to a slight underreporting of being overweight (Rowland 1989; cf. Gortmaker et al. 1993).
The U.S. Centers for Disease Control (CDC) specify four categories for adult BMI: (1) a BMI of less than 18.5 isunderweight, (2) between 18.5 and 25 isnormal weight, (3) between 25 and 30 is overweight, and (4) greater than 30 isobese(with subcategories of class I obesity [BMI = 30 to 35], class II obesity [BMI = 35 to 40], and class III obesity [BMI > 40]). Following Gortmaker and asso- ciates (1993), I tested income differences between those who were“obese”(BMI≥30) and all others. For each iteration of the statistical model, I also tested differences between those who could be classified as“class II/III obese”(BMI≥35) and all others. This attention to variation in the obese population is supported by recent literature (Carr and Friedman 2005) as well as by in- creases in the average weight of the U.S. population in the past 20 years. A number of studies have shown that people’s perception of their own and others’weight is constructed relative to their so- cial context (Chang and Christakis 2003; Perrin, Flower, and Ammerman 2005), and so it seems possible that, in the context of greater average weights in the population, the BMI threshold for anti-fat discrimination might rise as well.
Finally, I opted to use height and weight datafrom 2001. Although BMI data from 2008 were available, using the 2001 data allows me to follow the pattern established by Gortmaker and associ- ates (1993) and Averett and Korenman (1996) of using heights and weights lagged by seven years.
Critically, using lagged weight more clearly establishes the direction of causation between weight (in late adolescence, in this case) and later wage outcomes, and it decreases the chance of the association between weight and income being due to some confounding factor. I examined the relationship be- tween 2001 and 2008 BMI and found a correlation of .76, with women gaining an average of 17 pounds (and 3 BMI points) over the seven elapsed years and men gaining 20 (and 3 BMI points).
Educational Attainment.One of the most important predictors of income is educational attain- ment, so I included a variable for education early on in the statistical models. Gortmaker and as- sociates (1993) tested both a year-by-year continuous variable for education and a binary variable for whether respondents had completed college or not. I did the same, based on respondents’ed- ucational attainment as of 2008. However, I will only display results of the binary variable (“Did the respondent complete college or not?”) in this article, as those models consistently showed greater explanatory power vis-à-vis 2008 income than did the year-by-year education variable I tested. This finding may indicate the increasing necessity of having a college degree to succeed in the job market in the late 2000s.
Parenthood.Having children can significantly affect one’s career choices and opportunities. For the purposes of this research, either biological parenthood or adopting a child caused respondents to be labeled as parents. Further, the effects of parenthood may differ by gender, since motherhood itself may sometimes be a basis for discrimination (along with the weight gains often accompanying childbirth), and fatherhood is sometimes associated with raised incomes (through an informal “family wage”or“breadwinner’s bonus”) (Correll, Benard, and Paik 2007; Waldfogel 1998).
Work Experience.The NLSY97 allows us to measure respondents’work experience by asking, for every year since the start of the survey, how many weeks the respondent has been employed during the year. For the purposes of my research question, I operationalized work experience as the total number of weeks respondents reported working since turning 18.
I also created a squared term to account for any curvilinear tendencies of the work experi- ence variable. It seemed likely that, similar to the effects of age on income late in life, the benefits associated with incremental increases in work experience (and decreases in schooling) might level off or even start to decline at some point. 422MASON Occupation.In light of Carr and Friedman’s (2005) finding of a significant effect of occupation- al status on experiencing weight-based discrimination, I included a variable to indicate whether or not the respondent was employed in an upper white-collar occupation in 2008. My classifica- tion of occupations as upper white-collar (management, professions) or not (clerical staff, service workers, blue-collar occupations, and farmers) is based on Carr and Friedman’s operationalization of this variable, and falls in line with recommendations from the U.S. Census Bureau.
Interaction Terms.Finally, I constructed a number of interaction terms indicated by the litera- ture and to test for statistical discrimination:obese and work experience(both obese x work experi- ence and obese x work experience 2);obese and occupation;andobese and race/ethnicity(with obese multiplied by each of the dummy variables for race/ethnicity, with white as the reference category).
Control Variables.I included a set of additional variables in later versions of the linear regres- sion model to account for other possible causes of income inequality. Gender was already ac- counted for because I ran separate regressions for men and women, following the work of Gortmaker and associates (1993) and because the literature indicated that certain control varia- bles might have opposite effects for women and men. In terms of racial/ethnic identity, the NLSY97 included separate variables for race (white, black, Native American, Asian, or other) and ethnicity (Hispanic or non-Hispanic). Race/ethnicity was then recoded as a series of binary dum- my variables: white non-Hispanic, black non-Hispanic, Hispanic, and non-Hispanic other (this last group made up such a small percentage of the total observations, about 3.5 percent, that any fur- ther divisions would have made it hard to draw statistically significant conclusions).
Following Gortmaker and associates’(1993) model, I included a series of family background variables: household income per capita in 1997, years of education the respondent’sfatherhad completed, and years of education the respondent’s mother had completed.
Finally, I included a number of variables that dealt with the respondent’s own capabilities for work according to Gortmaker and associates’formulation. The 2008 questionnaire asked whether respondents had any health conditions that limited either the quantity or type of work they were able to do. Further, the 1999 version of the NLSY97 administered the Armed Services Vocational Aptitude Battery (ASVAB) to participants, which is roughly equivalent to the Armed Forces Qual- ifying Test (AFQT) used in the original NLSY. Both tests are meant to provide an approximate measure of intellectual ability. Scores on the ASVAB ranged between 0 and 100, with average scores of 54.5 for women and 52.6 for men. Procedures To test the relationship between 2001 weight and 2008 income, I first ran a bivariate ordinary least squares (OLS) regression between logged 2008 income and 2001 weight (obese or nonob- ese) using the sample weights provided by NLSY97. Following Gortmaker and associates’exam- ple, I ran each model separately for men and women (Model 1a), a decision that allowed me to capture differing effects of gender on both key variables and controls in subsequent models. Next, I added variables for the respondent’s education level and parental status (Model 2a), work experi- ence (both a linear term and a squared term) (Model 3a), occupation (Model 4a), and then a series of control variables (Model 5a). Last, I tested the three interactions—obese x work experience, obese x occupation, and obese x race/ethnicity—in the final model (Models 6a, 7a, and 8a). I then repeated these steps to test the effects of class II/III obesity on income (Models 1 through 8b).
Regression coefficients from the different models are shown in Tables 2 through 5. Unless other- wise specified, reported results will refer to the full basic model with the lower BMI cutoff for obesity, Model 5a (Table 2). The Unequal Weight of Discrimination423 Results Key Variables For women, being obese (BMI > 30) proved to be a significant predictor of lowered income.
This finding indicates that the ties between income and body size continued to be important fac- tors influencing women’s life chances in the 2000s, extending into the twenty-first century the findings that Gortmaker and associates (1993) and other authors published on data from the 1980s. Holding all other variables constant at their averages, obese women earned an average of $15,220 a year while nonobese women earned an average $18,948. On the other hand, obesity was associated with smaller negative coefficients for men than for women, and male obesity had no significant effect on income in the full model (Model 5a, Table 2).
The gender dynamics of the weight-income relationship shift when obesity is redefined as including only those with BMI scores higher than 35 (see Table 3). For women, the significance and magnitude of the obesity coefficient remain virtually unchanged, such that obese II/III wom- en made an average of $14,592/year relative to the $18,659 earned by their thinner peers. For men, membership in the obese II/III category becomes a significant negative predictor of income:
those with BMIs of 35 or greater earned an average of $16,166 per year, compared to the $25,406 earned by thinner men—in other words, lower-BMI men made 57 percent higher incomes per year. All told, these findings provide evidence that weight-based income penalties begin at lower weights for women than for men (providing partial support for H 3), but that very obese (class II/III) men are heavily penalized relative to nonobese men.
Work experience was a significant predictor of income for both women and men, whereas the polynomial (squared) formulation of work experience was significant for women only. For women workers, then, this finding suggests that incremental increases in experience during the early years of work yielded a much higher payoff per year than increased experience further down the line, whereas men in the labor force continued to increase their incomes as they gained work experience. A caveat to these work experience findings is that all of the subjects in this study were in the very early stages of their careers in 2008, and some of their past experience might have been in seasonal or temporary jobs unrelated to their career goals. If such trends continue over a longer period of time, though, it might suggest that men have an easier time getting their accomplishments and ability noticed and getting pay raises. Controls Women in this study received substantial benefits for having higher ability (as measured by the ASVAB), earning a .3 percent increase in salary for every additional point they scored on the test. Higher scores on the ASVAB did not translate into increased earnings for men, however, reaching only marginal significance in the full basic model. Likewise, attaining a college degree or higher contributed significantly to higher incomes for women (leading to a 17 percent salary bump), while it was not significant for men. Finally, working in an upper white-collar occupation was significantly associated with a 15 percent earnings premium for both women and men. Thus, when occupation is controlled for, women’s earnings appear to be much more variable due to meritocratic discrimination (based on education and ability) and to weight, whereas men see rel- atively little effect on their income due to education, ability, or—with the exception of very obese men—weight.
In addition to education and ability, two characteristics seen by many as more rational, mer- itocratic bases for income disparities, the regression also controlled for a number of characteristics that have been historically subjected to prejudicial discrimination.
Race/Ethnicity.After adding controls for relevant background variables, I found no significant racial/ethnic differences in income among women or men. These inconclusive findings about 424MASON Table 2Contributors to Income (logged, 2008) for U.S. Women and Men Aged 23–29 (BMI > 30, Models 1 through 5) Women Men Model 1a Model 2a Model 3a Model 4a Model 5a Model 1a Model 2a Model 3a Model 4a Model 5a R 2 .01 .08 .19 .20 .21 .00 .02 .14 .15 .18 Constant9.92*(.03)9.84*(.05)8.22*(.27)8.20*(.26)7.81*(.44)10.18*(.03)10.04*(.03)9.10*(.21)9.05*(.21)7.58*(.40) Obese (BMI > 30) .32*(.10) .22*(.10) .25*(.10) .25*(.10) .22*(.10)−.09 (.08)−.06 (.08)−.14 †(.08)−.13 †(.08)−.11 (.07) College degree (2008).35*(.06).31*(.06).24*(.06).17*(.07).24*(.06).26*(.06).17*(.06) .09 (.06) Have any children (2008) .23*(.06) .23*(.06) .23*(.06) .21*(.06).20*(.05).13*(.05).14*(.05).19*(.05) Weeks worked (2008).01*(.00).01*(.00).01*(.00).00*(.00).00*(.00).00*(.00) Weeks worked (2008) 2 .00*(.00) .00*(.00) .00*(.00)−.00 (.00)−.00 (.00)−.00 (.00) White collar worker (2008).16*(.06).15*(.06).17*(.05).15*(.05) Race Black.07 (.08)−.06 (.09) Hispanic.07 (.10) .05 (.06) Other.11 (.12) .17 (.14) Socioeconomic background 1997 household income (log) .03 (.04).15*(.03) Mother’s education−.02 (.01) .00 (.01) Father’s education .01 (.01)−.01 (.01) Personal characteristics Any health limitation (2008)−.04 (.12) .42*(.13) ASVAB (1999).003*(.00) .002 †(.00) Source: National Longitudinal Survey of Youth 97 (1997, 1999, 2001, and 2008 cases).
Note: Numbers in parentheses are the standard deviations associated with the coefficients directly above them.
*coefficient is significant at a 95% level of confidence. †coefficient is marginally significant at a 90% level of confidence. The Unequal Weight of Discrimination425 Table 3Contributors to Income (logged, 2008) for U.S. Women and Men Aged 23–29 (BMI > 35, Models 1 through 5) Women Men Model 1b Model 2b Model 3b Model 4b Model 5b Model 1b Model 2b Model 3b Model 4b Model 5b R 2 .00 .07 .19 .20 .20 .01 .03 .15 .16 .19 Constant9.90*(.03)9.82*(.05)8.20*(.27)8.19*(.26)7.74*(.44)10.18*(.02)10.06*(.03)9.12*(.21)9.07*(.21)7.60*(.39) ObeseII/III(BMI>35) .33*(.12)−.20 †(.12) .27*(.12) .27*(.12) .25*(.12) .50*(.22) .45*(.22) .51*(.20) .51*(.20) .45*(.18) College degree (2008).36*(.06).32*(.06).25*(.07).17*(.07).23*(.06).24*(.06).16*(.06) .07 (.06) Have any children (2008) .24*(.06) .23*(.06) .23*(.06) .21*(.06).19*(.05).12*(.05).14*(.05).18*(.05) Weeks worked (2008).01*(.00).01*(.00).01*(.00).00*(.00).00*(.00).00*(.00) Weeks worked (2008) 2 .00*(.00) .00*(.00) .00*(.00)−.00 (.00)−.00 (.00)−.00 (.00) White collar worker (2008).16*(.06).15*(.06).17*(.05).16*(.05) Race Black.06 (.07)−.05 (.09) Hispanic.07 (.10) .05 (.06) Other.12 (.12) .18 (.13) Socioeconomic background 1997 household income (log) .04 (.04).14*(.03) Mother’s education−.02 †(.01) .01 (.01) Father’s education .01 (.01)−.01 (.01) Personal characteristics Any health limitation (2008)−.06 (.12) .39*(.12) ASVAB (1999).003*(.00) .002 †(.00) Source: National Longitudinal Survey of Youth 97 (1997, 1999, 2001, and 2008 cases).
Note: Numbers in parentheses are the standard deviations associated with the coefficients directly above them.
*coefficient is significant at a 95% level of confidence. †coefficient is marginally significant at a 90% level of confidence. 426MASON Table 4Contributors to Income (logged, 2008) for U.S. Women and Men Aged 23–29 (BMI > 30, Models 6 though 8) Women Men Model 6a Model 7a Model 8a Model 6a Model 7a Model 8a R 2 .21 .21 .21 .18 .18 .18 Constant7.85*(.46)7.80*(.44)7.82*(.44)7.61*(.40)7.57*(.40)7.58*(.40) Obese (BMI > 30)−.45 (.50)−.08 (.09) .32*(.13)−.46 .33−.12 (.08)−.15 (.09) College degree (2008).17*(.07).17*(.07).17*(.07) .08 (.06) .09 (.06) .09 (.06) Have any children (2008) .22*(.06) .21*(.06) .21*(.06).18*(.05).19*(.06).19*(.05) Weeks worked (2008).01*(.00).01*(.00).01*(.00).00*(.00).00*(.00).00*(.00) Weeks worked (2008) 2 .00*(.00) .00*(.00) .00*(.00)−.00 (.00)−.00 (.00)−.00 (.00) White collar worker (2008).16*(.06).19*(.06).16*(.06).15*(.05).14*(.05).15*(.05) Race Black .08 (.08) .06 (.07) .03 (.08)−.07 (.09)−.07 (.09)−.08 (.10) Hispanic .06 (.10) .06 (.10) .01 (.10) .05 (.06) .05 (.06) (.03) (.06) Other .11 (.12) .11 (.12) .12 (.13) .18 (.14) .17 (.14) .09 (.17) Socioeconomic background 1997 household income (log) .03 (.04) .04 (.04) .03 (.04).14*(.03).15*(.03).15*(.03) Mother’s education−.02 (.01)−.02 (.01)−.02 (.01) .01 (.01) .00 (.01) .00 (.01) Father’s education .01 (.01) .01 (.01) .01 (.01)−.01 (.01)−.01 (.01)−.01 (.01) Personal characteristics Any health limitation (2008)−.04 (.12)−.04 (.12)−.04 (.12) .41*(.13)−.42* (.13) .41*(.13) ASVAB (1999).003*(.00).003*(.00).003*(.00) .001 (.00) .002 †(.00) .002 †(.00) Interaction effects Obese and weeks worked .00 (.00) .00 (.00) Obese and weeks worked 2 .00 (.00)− Obeseandwhitecollar−.33 (.21) .06 (.20) Obese and black .28 (.19) .08 (.19) Obese and Hispanic.68*(.20) .15 (.15) Obese and other−.06 .35 −.36−.26 Source: National Longitudinal Survey of Youth 97 (1997, 1999, 2001, and 2008 cases).
Note: Numbers in parentheses are the standard deviations associated with the coefficients directly above them.
*coefficient is significant at a 95% level of confidence. †Coefficient is marginally significant at a 90% level of confidence. The Unequal Weight of Discrimination427 Table 5Contributors to Income (logged, 2008) for U.S. Women and Men Aged 23–29 (BMI > 35, Models 6 though 8) Women Men Model 6b Model 7b Model 8b Model 6b Model 7b Model 8b R 2 .20 .20 .21 .20 .19 .19 Constant7.74*(.44)7.75*(.44)7.71*(.44)7.68*(.39)7.60*(.38)7.61*(.38) Obese II/III (BMI > 35)−.08 (.70)−.15 (.12)−.24 (.15) 1.64*(.60) .39*(.18) .61*(.26) College degree (2008).17*(.07).18*(.07).17*(.07) .07 (.06) .07 (.06) .08 (.06) Have any children (2008) .21*(.06) .21*(.06) .21*(.06).17*.05.18*.05.18*.05 Weeks worked (2008).01*(.00).01*(.00).01*(.00).00*(.00).00*(.00).00*(.00) Weeks worked (2008) 2 .00*(.00) .00*(.00) .00*(.00)−.00 (.00)−.00 (.00)−.00 (.00) White collar worker (2008).15*(.06).16*(.06).16*(.06).15*(.05).17*(.05).16*(.05) Race Black .06 (.07) .06 (.07) .05 (.08)−.05 (.09)−.05 (.09)−.08 (.09) Hispanic .07 (.10) .07 (.10) .07 (.10) .04 (.06) .05 (.06) .03 (.06) Other .12 (.12) .12 (.12) .15 (.12) .16 (.13) .17 (.13) .17 (.14) Socioeconomic background 1997 household income (log) .04 (.04) .04 (.04) .04 (.04).14*(.03).14*(.03).15*(.03) Mother’s education−.02 †(.01)−.02 †(.01)−.02 (.01) .00 (.01) .01 (.01) .01 (.01) Father’s education .01 (.01) .01 (.01) .01 (.01)−.01 (.01)−.01 (.01)−.02 (.01) Personal characteristics Any health limitation (2008)−.06 (.12)−.05 (.12)−.06 (.12) .34*(.12) .39*(.12) .39*(.12) ASVAB (1999).003*(.00).003*(.00).003*(.00) .002 †(.00) .001 (.00) .002 (.00) Interaction effects Obese and weeks worked .00 (.00).00*(.00) Obese and weeks worked 2 −.00 (.00)– Obeseandwhitecollar−.24 (.25)−.24 (.50) Obese and black .15 (.23) .52 †(.31) Obese and Hispanic .02 (.34) .59 (.37) Obese and other .70*−.36 .13 (.35) Source: National Longitudinal Survey of Youth 97 (1997, 1999, 2001, and 2008 cases).
Note: Numbers in parentheses are the standard deviations associated with the coefficients directly above them.
*Coefficient is significant at a 95% level of confidence. †Coefficient is marginally significant at a 90% level of confidence. 428MASON race/ethnicity and weight may have been due, in part, to relatively small sample sizes for the various racial/ethnic groups I studied. Additionally, including aptitude test scores (such as the ASVAB that study participants took) in a regression can mask the differential effects of race, since such tests may privilege certain cultural (white, class privileged) knowledge and Anglophone English. I did, however, find some significant and marginally significant effects of interactions between obesity and various racial/ethnic categories, which I will discuss in the next section.
Socioeconomic Background (logged 1997 household income).For men, logged 1997 household income, a rough estimator of respondents’childhood SES, was positively and significantly related to 2008 income. For women, however, family socioeconomic background was unimportant, and for both men and women, neither fathers’nor mothers’educational level in 1997 had a significant effect on 2008 income.
Parenthood.Other interesting gender differences can be seen in the effects on income when respondents became parents themselves. Working women with at least one child faced a penalty of 21 percent less income than childless women. For men, on the other hand, being a father was associated with anincreaseof 19 percent per year, even independent of age, race/ethnicity, or family background. This perhaps represents choices that heterosexual couples make in child care arrangements and placing one partner’s career (usually the man’s) ahead of that of the other, and it may imply gaps in women’s employment histories. Also, these findings support Shelley Correll, Stephen Benard, and In Paik’s (2007) thesis that motherhood, as a status charac- teristic, comes with a wage penalty for women, while men receive an informal“breadwinner’s bonus.” Health.Although some have posited that health conditions may mediate the relationship be- tween fatness and lowered incomes (most notably, the U.S. Centers for Disease Control [CDC 2010]), my analysis found that health limitations on subjects’ability to work only had a significant (negative) impact on men’s incomes (such limitations were associated with 42 percent lower in- comes), while women’s incomes were unaffected by medical conditions. It is crucial to remember that health limitations were self-reported by respondents, and that they representedactualhealth problems to date, not potential ones. Therefore, although the obese women of this study did not suffer from diminished earnings due to actual ill health, it is still possible that employers perceived them as beingat riskfor illness. Having a legitimate financial interest in healthy employees due to concerns about both productivity and liability for health care costs under the United States’s cur- rent employer-based health care system, a manager might pass over a healthy worker whom she or he believed to be at a higher risk of disease. Were this to be the case, it would represent an in- stance of statistical discrimination.
Obese I-III versus Obese II/III.As noted earlier, using a BMI cutoff of 30 (Model 5a) versus 35 (Model 5b) resulted in measurably different findings for the gendered effects of weight on income.
Apart from the effects of weight itself, though, the coefficients for all other independent variables were virtually identical across the two models. Interactions Obesity and Work Experience.The interaction between obesity and work experience, which forms the basis of the test for statistical discrimination, was not significant for women or men when obesity was set at its lower cutoff point (BMI > 30) (see Table 4). Likewise, the model using a higher cutoff point for obesity (BMI > 35) still showed no significant interaction effect between women’s weight and their work experience (see Table 5).This finding suggests that any effect of obesity on women’s income was not due to statistical discrimination and may not have been helped by persistent hard work and visibility in the workplace.Thus, this finding is very suggestive of prejudicial discrimination The Unequal Weight of Discrimination429 (H2) against obese women, although such a conclusion would appear to be based in a residual explanation of discrimination.
In contrast, the effects of male obesity (class II/III) on income strongly indicated statistical dis- crimination against very obese men (H 1), due to the fact that the interaction effect between work experience and obesity was significant (see Table 5). In general, nonobese men’sincomestended to lie well above women’s, and obese men’s incomes were either higher than all women’s(inthe case of class I-III obese men, Model 6a) or predicted to overtake women after several years’work experience (in the case of class II/III obese men, Model 6b). Figure 1 shows the different effects of work experience on income for men and women, differentiated by whether or not they were obese (class I-III), while Figure 2 shows the same effects differentiated by whether or not respondents were very obese (class II/III).
Obesity and Occupation.For both women and men, occupational type did not significantly in- teract with weight at either cutoff for obesity (BMI > 30 or BMI > 35). Although Carr and Fried- man (2005) found that obese white-collar workers were more likely than their nonobese colleagues to subjectively report experiencing discrimination in some form, the current data do not support that finding when it comes to objective measures of income discrimination. Further studies looking at workers with greater labor force experience and/or at other measures of em- ployment inequality may be needed to further substantiate Carr and Friedman’sfindings.
Obesity and Race/Ethnicity.Some racial/ethnic interactions with obesity were significant for women in the models I tested: Hispanic women were less likely to suffer income inequality due to being obese (BMI > 30) than white women, and—more ambiguously—women in the“other” racial/ethnic category were more likely than white women to face income inequality due to being 10.8 10.4 9.9 9.4 8.9 Predicted Log Income in 2008 8.4 100 200 300 Job Experience - Weeks Worked Since 18400 500 Under-, Normal, and Overweight Women (2001) Under-, Normal, and Overweight Men (2001) Obese I-III Women (2001) Obese I-III Men (2001) Figure 1Predicted 2008 Incomes for Men and Women (BMI > 30) Source: National Longitudinal Survey of Youth 97 Note: Obese is defined as BMI > 30. 430MASON very obese (BMI > 35). Men in both models showed no significant differences in the effects of obe- sity due to race/ethnicity. These findings align with those of other researchers, including Dalton Conley and Rebecca Glauber (2007) and Averett and Korenman (1996), who found that white women were more disadvantaged by weight-based discrimination than women of color (but that no such pattern existed for men), and with the statement by Carr and Friedman (2005) that“obesity is considered a greater normative violation among women, whites, and young persons”(p. 255). Discussion Past research has provided evidence for the existence of income discrimination against fat in- dividuals. This study’s findings suggest that obese women and men experience such discrimina- tion differently, in terms of both duration and intensity. Through examination of an interaction effect between work experience and obesity, it appears that very obese men in this study were economically disadvantaged in the early years of work relative to their more normative-bodied, nonobese peers, but over time, they were able to make up the difference. This finding suggests that their early lowered wages may have been due to statistical discrimination—assumptions made by employers that eventually could be corrected with time spent on the job.
Obese women, on the other hand, earn less than their thinner female counterparts at every step of their careers, a conclusion that is supported by my own findings as well as the existing literature. For example, Conley and Glauber’s (2007) study of body mass effects on later life 10.8 10.4 9.9 9.4 8.9 Predicted Log Income in 2008 8.4 100 200 300 Job Experience - Weeks Worked Since 18400 500 Under-, Normal, Overweight and Obese I Women (2001) Under-, Normal, Overweight and Obese I Men (2001) Obese II/III Women (2001) Obese II/III Men (2001) Figure 2Predicted 2008 Incomes for Men and Women (BMI > 35) Source: National Longitudinal Survey of Youth 97 Note: Obese is defined as BMI > 35. The Unequal Weight of Discrimination431 socioeconomic status found that income penalties for obese women in 2001 persisted into their mid-fifties, but that no such effect existed for obese men. Clearly, there is a negative relationship between body mass and income for women, and it does not disappear with time on the job. Hav- ing thus ruled out H 0(no relationship) and H 1(statistical discrimination), the most likely explana- tion for obese women’s income disadvantage is prejudicial discrimination.Further, this study lends support to previously suggested differences in the effects of being fat for men and for women (H 3). Specifically, there appears to be a gender gap not only in men and women’s wages, but also in the degree to which they are disadvantaged by having nonnormative bodies. Women in the NLSY97 faced severe income penalties for being obese, and these effects’significance and magni- tude remained robust in the presence of a wide variety of control variables. Such direct income effects represent only one of several mechanisms through which fat women have been found to suffer limited access to economic and social resources. Indirect causes of fat women’s economic disadvantage include, first of all, lower levels of education, due in part to educational“cooling- out”processes that encourage them to lower their sights. Marilyn Wann (2009) describes research by C. S. Crandall (1995) showing that“[h]igh school counselors are less likely to encourage fat students to apply for college, colleges are less likely to admit equally qualified fat applicants, and parents are less likely to pay a fat daughter’s college tuition”(p. xix). A second indirect cause of economic hardship for fat women is that they face worse marriage prospects, in terms of both their lower likelihood of ever marrying and the lower incomes of the men they do marry (Averett and Korenman 1996; Gortmaker et al. 1993). Thus, fat women—perhaps similar to the situation of women in other marginalized groups—are less able to rely on economic support provided by a husband, but at the same time they are less encouraged to develop the skills needed to support themselves. My findings in the realm of income inequality suggest that obese women who over- come these obstacles and find employment will be compensated less for their efforts than other women, not more. Finally, while the regression models in this study control for a variety of back- ground factors, it is worth noting that weight-based income inequality for women may compound a number of other factors that work to further disadvantage obese women: the obese women in my sample were significantly less likely to have completed college, significantly less likely to be married, significantly more likely to have a child, and from childhood homes with significantly lower incomes than those of their nonobese female peers.
Although I found statistical discrimination (in the form of a significant interaction between weight and work experience) only for very obese men, the different patterns of weight-based wage gaps among women and men help to shed light on some of the mechanisms underlying this inequality. As noted at the beginning of this article, one popular explanation of wage gaps between fat and slender workers is that differences in pay are attributable to actual differences in productivity or ability (i.e., meritocratic discrimination). Controlling for education, experience, in- tellectual ability, and other factors can help make this explanation less viable, but it is still not pos- sible to completely rule it out. There might be some as-yet-unaccounted-for variable that would explain the persistent link between higher weight and lower pay.
The finding of gender differences is instructive in this regard. For men, I have shown that sta- tistical discrimination was taking place; very obese men were indeed disadvantaged, but they could overcome initial perceptions and close the wage gap over time. Though there was no signif- icant interaction effect to suggest statistical discrimination for women, the men’s findings can nonetheless help us understand the nature of the disadvantages women faced. If obese men, over time, were able to work hard and prove their worth, then the low pay they faced early on was due to incorrect employer perceptions,notto some unobserved physical or mental deficiency evi- denced by their bodies.
If inferior physical or mental abilities were not the cause of obese men’s disadvantages, then it seems unlikely that they would be the cause of obese women’s disadvantages, either. Why would obese women suffer from some pathology affecting their ability to work, whereas obese men would not? A more likely explanation is that obese women and men both suffered from negative perceptions by employers. The gender difference, then, would result from the different degrees to 432MASON which women and men continued to be judged based on their appearances. In a society where sexist attitudes still hold sway, women are expected to demand fewer resources and take up less space—to be physically less than men (Bordo [1993] 2003). Furthermore, a pervasive cultural be- lief in sexual dimorphism (the notion that there are absolute, distinct physical differences between all men and all women) does more than describe the average bodily differences between women and men (e.g., that men are bigger and more muscular than women), it also disciplines those bod- ies in order to (re)produce sexual dimorphism. Thus, while both obese women and men experi- ence a variety of disadvantages (in this case, income inequality) due to the stigmatization of fatness itself, obese women are further penalized for failing to do their gender properly: in short, their fatness causes them to be punished for being unfeminine. Thus, not only should studies on the social significance of fatness pay special attention to gender, but also, I argue, gender scholar- ship dealing with intersectionality should consider body size as a vector of inequality that, like race/ethnicity or class, interacts with and mediates gender. Conclusion This article contributes to the current literature in three key respects. First, previous studies attempting to statistically prove the existence of anti-fat income discrimination have often done so by pointing to residual income differences after controlling for a number of explanatory varia- bles. Yet, one popular counter-argument to these studies claims that wage gaps between fat and slender workers are attributable to actual differences in productivity or ability (i.e., meritocratic discrimination). Controlling for education, experience, intellectual ability, and other factors can help make this explanation less viable, but can never fully eliminate it. By first theorizing a spec- trum of discrimination ranging from lawful meritocratic discrimination to illegal statistical and prejudicial discrimination, then introducing Altonji and Blank’s (1999) test for statistical discrimi- nation, this article provides positive evidence for the existence of weight-based employment dis- crimination.
Second, most of the existing studies on weight-based income inequality draw on data from the 1980s and 1990s. To my knowledge, this article is the first to investigate income inequality and body size using the NLSY97 cohort, and one of the first to look at these trends in the late 2000s.
Taken together, the present study and past work provide compelling evidence of the decades-long persistence of weight-based employment discrimination in the United States. To the extent that public policy engages with these findings, it tends to pose individual weight loss as the solution, rather than legal protections against anti-fat discrimination. However, given widespread evidence that weight-loss diets do not provide long-term weight reduction or health benefits for the major- ity of people (as in the 31 weight loss studies that Traci Mann and colleagues [2007] scrutinized in their meta-analysis), individual weight loss is not a viable solution to the ongoing problem of weight-based income inequality in the United States that this study illustrates.
Third, this study provides evidence of important gender differences in how women and men experience weight-based discrimination. For women, anti-obesity discrimination sets in at lower body mass levels and does not dissipate over time as they gain experience. This income penalty stacks on top of other economic disadvantages past research has found for this population, includ- ing lower average educational attainment and marriage rates, and it may be more pronounced for white women than for certain racial and ethnic minority women. For men, obesity may be linked to lowered income relative to thinner male colleagues, but only among those who are very obese (BMI > 35), a cutoff Carr and Friedman (2005) describe as the point at which“obesity may be- come a‘master status,’or a characteristic that overrides all other features of a person’s identity” (p. 255). Yet even in this case, obese men’s disadvantage is not as severe as women’s. Obese men can overcome initial disadvantages with time and experience, whereas obese women will contin- ue to face diminished wages over the course of a lifetime. The Unequal Weight of Discrimination433 References Altonji, Joseph G. and Rebecca M. Blank. 1999.“Race and Gender in the Labor Market.”Pp. 3143–259 in Handbook of Labor Economics,Vol. 3, edited by Orley Ashenfelter and David Card. Amsterdam: Elsevier Science B.V.
Averett, Susan and Sanders Korenman. 1996.“The Economic Reality of the Beauty Myth.”Journal of Human Resources31:304–30.
Baum, Charles and William Ford. 2004.“The Wage Effects of Obesity: A Longitudinal Study.”Health Economics 13(9):885–99.
Becker, Gary S. 1975.Human Capital. 2d ed. New York: National Bureau of Economic Research.
Berger, John. [1972] 1977.Ways of Seeing. New York: Penguin Books.
Biddle, Jeff E. and Daniel S. Hamermesh. 1998.“Beauty, Productivity, and Discrimination: Lawyers’Looks and Lucre.”Journal of Labor Economics16(1):172–201.
Boero, Natalie. 2007.“All the News that’s Fat to Print: The American‘Obesity Epidemic’and the News Media.” Qualitative Sociology30(1):41–60.
Bordo, Susan. [1993] 2003.Unbearable Weight: Feminism, Western Culture, and the Body. Berkeley: University of California Press.
Bourdieu, Pierre. [1979] 1984.Distinction: A Social Critique of the Judgement of Taste. Translated by Richard Nice.
Cambridge, MA: Harvard University Press.
Butler, Judith. [1990] 2006.Gender Trouble: Feminism and the Subversion of Identity. New York: Routledge.
Carr, Deborah and Michael A. Friedman. 2005.“Is Obesity Stigmatizing? Body Weight, Perceived Discrimina- tion, and Psychological Well-Being in the United States.”Journal of Health and Social Behavior46(3):
244–59.
Cawley, John. 2004.“The Impact of Obesity on Wages.”Journal of Human Resources39:451–74.
Cawley, John and Sheldon Danziger. 2005.“Morbid Obesity and the Transition from Welfare to Work.” Journal of Policy Analysis and Management24(4):727–43.
CDC. 2010.“Overweight and Obesity: Economic Consequences.”Retrieved January 31, 2010 (www.cdc.gov/ obesity/causes/economics.html).
Chang, V. W. and N. A. Christakis. 2003.“Self-Perception of Weight Appropriateness in the United States.” American Journal of Preventative Medicine24(4):332–39.
Conley, Dalton and Rebecca Glauber. 2007.“Gender, Body Mass, and Socioeconomic Status: New Evidence from the PSID.”Advances in Health Economics and Health Services Research17:253–75.
Correll, Shelley J., Stephen Benard, and In Paik. 2007.“Getting a Job: Is There a Motherhood Penalty?” American Journal of Sociology112:1297–1338.
Crandall, C. S. 1995.“Do Parents Discriminate Against their Heavyweight Daughters?”Personality and Social Psychology Bulletin21(7):742–35.
Durkheim, Emile. [1933] 1997.The Division of Labour in Society. Edited by Lewis Coser. New York: The Free Press.
Flegal, Katherine M., Barry I. Graubard, David F. Williamson, and Mitchell H. Gail. 2005.“Excess Deaths Associated with Underweight, Overweight, and Obesity.”Journal of the American Medical Association 293(15):1861–67.
Gortmaker, Steven L., Aviva Must, James M. Perrin, Arthur M. Sobol, and William H. Dietz. 1993.“Social and Economic Consequences of Overweight in Adolescence and Young Adulthood.”New England Journal of Medicine329(14):1008–12.
Hamermesh, Daniel S. and Jeff E. Biddle. 1994.“Beauty and the Labor Market.”American Economic Review 84(5):1174–94.
Haskins, Katherine and H. Edward Ransford. 1999.“The Relationship between Weight and Career Payoffs among Women.”Sociological Forum14(2):295–318.
Kristen, Elizabeth. 2002.“Addressing the Problem of Weight Discrimination in Employment.”California Law Review90(1):57–109.
Mann, Traci, Janet Tomiyama, Erika Westling, Ann-Marie Lew, Barbra Samuels, and Jason Chatman. 2007.
“Medicare’s Search for Effective Obesity Treatments: Diets Are Not the Answer.”American Psychologist 62(3):220–33.
McDowell, Margaret A., Cheryl D. Fryar, Cynthia L. Ogden, and Katherine M. Flegal. 2008.“Anthropometric Reference Data for Children and Adults: United States, 2003–2006.”National Health Statistics Reports 10.
Retrieved February 16, 2012 (www.cdc.gov/nchs/data/nhsr/nhsr010.pdf). 434MASON Medoff, James L. and Katharine G. Abraham. 1981.“Are Those Paid More Really More Productive? The Case of Experience.”Journal of Human Resources16(2):186–216.
Mulford, Matthew, John Orbell, Catherine Shatto, and Jean Stockard. 1998.“Physical Attractiveness, Oppor- tunity, and Success in Everyday Exchange.”American Journal of Sociology103(6):1565–92.
Mulvey, Laura. 1975.“Visual Pleasure and Narrative Cinema.”Screen16(3):6–18.
Ortner, Sherry. 1974.“Is Female to Male as Nature Is to Culture?”Pp. 68–87 inWoman, Culture, and Society, edited by Michelle Rosaldo, Louise Lamphere, and Joan Bamberger. Stanford, CA: Stanford University Press.
Pace, Gina. 2006.“Obesity Bigger Threat than Terrorism?”Associated Press, March 1. Retrieved August 9, 2010 (www.cbsnews.com/stories/2006/03/01/health/main1361849.shtml).
Perrin, Eliana M., Kori B. Flower, and Alice S. Ammerman. 2005.“Pediatricians’Own Weight: Self-Perception, Misclassification, and Ease of Counseling.”Obesity Research13(2):326–32.
Puhl, Rebecca M. and Chelsea A. Heuer. 2010.“Obesity Stigma: Important Considerations for Public Health.” American Journal of Public Health100(6):1019–28.
Puhl, Rebecca M. and Kelly D. Brownell. 2001.“Bias, Discrimination, and Obesity.”Obesity Research9(12):
788–805.
Puhl, Rebecca M., Tatiana Andreyeva, and Kelly D. Brownell. 2008.“Perceptions of Weight Discrimination:
Prevalence and Comparison to Race and Gender Discrimination.”International Journal of Obesity 32:992–1000.
Roehling, Mark V. 2002.“Weight Discrimination in the American Workplace: Ethical Issues and Analysis.” Journal of Business Ethics40(2):177–89.
Rowland, M. L. 1989.“Statistical Bulletin, Metropolitan Life Insurance Company.”New York: Metropolitan Life Insurance Company.
Saguy, Abigail C. 2008.“French Women Don’t Get Fat?”Health at Every Size Journal 19(4):219–34.
Saguy, Abigail C. and Kevin C. Riley. 2005.“Weighing Both Sides: Morality, Mortality, and Framing Contests over Obesity.”Journal of Health Politics, Policy, and Law30(5):869–921.
Schafer, Markus H. and Kenneth F. Ferraro. 2011.“The Stigma of Obesity: Does Perceived Weight Discrimina- tion Affect Identity and Physical Health?”Social Psychology Quarterly74(1):76–97.
Stunkard, Albert J. and Thorkild Sorensen. 1993.“Obesity and Socioeconomic Status–A Complex Relation.” New England Journal of Medicine329(14):1036–37.
Thurow, Lester C. 1975.Generating Inequality: Mechanisms of Distribution in the U.S. Economy. New York: Basic Books.
Waldfogel, Jane. 1998.“Understanding the‘Family Gap’in Pay for Women with Children.”Journal of Economic Perspectives12(1):137–56.
Wann, Marilyn. 2009.“Foreword–Fat Studies: An Invitation to Revolution.”Pp. ix-xxv inThe Fat Studies Reader, edited by Esther Rothblum and Sondra Solovay. New York: New York University Press.
Wolf, Naomi. [1991] 2002.The Beauty Myth. New York: HarperCollins.The Unequal Weight of Discrimination435
