american sociological review 2014 mcclintock 575 604

Upload: nodirtystuff

Post on 03-Jun-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    1/31

    http://asr.sagepub.com/AmericanSociological Review

    http://asr.sagepub.com/content/79/4/575Theonline version of this article can be foundat:

    DOI: 10.1177/0003122414536391

    2014 79: 575 originally published online 9 June 2014American Sociological ReviewElizabeth Aura McClintock

    Beauty and Status: The Illusion of Exchange in Partner Selection?

    Published by:

    http://www.sagepublications.com

    On behalf of:

    American Sociological Association

    can be found at:American Sociological ReviewAdditional services and information for

    http://asr.sagepub.com/cgi/alertsEmail Alerts:

    http://asr.sagepub.com/subscriptionsSubscriptions:

    http://www.sagepub.com/journalsReprints.navReprints:

    http://www.sagepub.com/journalsPermissions.navPermissions:

    What is This?

    - Jun 9, 2014OnlineFirst Version of Record

    - Jul 23, 2014Version of Record>>

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/content/79/4/575http://asr.sagepub.com/content/79/4/575http://asr.sagepub.com/content/79/4/575http://www.sagepublications.com/http://www.asanet.org/http://asr.sagepub.com/cgi/alertshttp://asr.sagepub.com/cgi/alertshttp://asr.sagepub.com/subscriptionshttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsPermissions.navhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://asr.sagepub.com/content/early/2014/05/29/0003122414536391.full.pdfhttp://asr.sagepub.com/content/79/4/575.full.pdfhttp://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://asr.sagepub.com/content/early/2014/05/29/0003122414536391.full.pdfhttp://asr.sagepub.com/content/79/4/575.full.pdfhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsReprints.navhttp://asr.sagepub.com/subscriptionshttp://asr.sagepub.com/cgi/alertshttp://www.asanet.org/http://www.sagepublications.com/http://asr.sagepub.com/content/79/4/575http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    2/31

    American Sociological Review2014, Vol. 79(4) 575604 American SociologicalAssociation 2014DOI: 10.1177/0003122414536391http://asr.sagepub.com

    This article revisits the claim that individuals

    (generally women) of relatively high physical

    attractiveness barter their beauty to attract a

    partner of higher socioeconomic status. This

    beauty-status exchange model is popularized in

    the trophy wife stereotype that pretty women

    marry high-status men. Yet this popular focus

    overlooks the role of matchingselecting a

    partner with similar characteristics to oneself

    that is well-documented in research on relation-ships (e.g., McPherson, Smith-Lovin, and Cook

    2001). I unite the two literatures by modeling

    exchange and matching as simultaneous and

    competing processes that might vary in

    relevance for different types of couples. Identi-

    fying the conditions under which couples

    exchange beauty and status provides insight

    into processes of social mobility and stratifica-

    tion. Additionally, beauty-status exchange is

    highly relevant to gender inequality and to

    sociobiological models of partner selection.

    ASRXXX10.1177/0003122414536391AmericanSociological ReviewMcClintock

    a

    University of Notre Dame

    Corresponding Author:Elizabeth Aura McClintock, 810 Flanner Hall,University of Notre Dame, Notre Dame, IN 46556E-mail: [email protected]

    Beauty and Status: TheIllusion of Exchange in

    Partner Selection?

    Elizabeth Aura McClintocka

    Abstract

    Scholars have long been interested in exchange and matching (assortative mating) in romanticpartner selection. But many analyses of exchange, particularly those that examine beautyand socioeconomic status, fail to control for partners tendency to match each other onthese traits. Because desirable traits in mates are positively correlated between partners andwithin individuals, ignoring matching may exaggerate evidence of cross-trait beauty-statusexchange. Moreover, many prior analyses assume a gendered exchange in which womentrade beauty for mens status, without testing whether men might use handsomeness toattract higher-status women. Nor have prior analyses fully investigated how the prevalenceof beauty-status exchange varies between different types of couples. I use data from theNational Longitudinal Study of Adolescent Health Romantic Pair Sample, a large (N= 1,507),nationally representative probability sample of dating, cohabiting, and married couples, toinvestigate how often romantic partners exchange physical attractiveness and socioeconomicstatus, net of matching on these traits. I find that controlling for matching eliminates nearly allevidence of beauty-status exchange. The discussion focuses on the contexts in which beauty-status exchange is most likely and on implications these results have for market-based andsociobiological theories of partner selection.

    Keywords

    marriage, demography, gender, social stratification, sociobiology

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    3/31

    576 American Sociological Review79(4)

    Partner selection is important in reproduc-

    ing group boundaries and social inequalities

    (Gordon 1964; Rosenfeld 2008). Matching on

    socioeconomic status compounds inequality:

    high-earners partner with other high-earnersand low-earners with other low-earners

    (Blossfeld and Buchholz 2009; Schwartz

    2010; Schwartz and Mare 2005). However,

    under a market model of partner selection, in

    which choices are relatively unconstrained by

    group boundaries and partners informally

    exchange personal attributes, individuals

    could leverage a desirable noneconomic trait

    to achieve socioeconomic mobility. This

    would undermine the tendency for partnerselection to reproduce existing patterns of

    inequality. Under the beauty-status exchange

    model, physical attractiveness might enable

    class mobility for women, although such an

    exchange would ensure womens economic

    dependency on their husbands.

    Indeed, partner selection is closely related

    to gender inequality. Norms dictating that the

    man be older and more successful in his

    career reinforce power inequalities withinmarriage (Presser 1975). Rules prescribing

    these and other forms of marital hypergamy

    may be weakening, but related stereotypes

    that men prioritize a partners appearance and

    women prioritize a partners status remain

    strong. Such stereotypes are pre-feminist,

    ignoring womens economic independence

    and their valuation of mens physical attrac-

    tiveness. Although these stereotypes may be

    derived from both social structural and socio-

    biological theories, the sociobiological litera-

    ture is especially adamant that men will select

    long-term partners on the basis of youth and

    beauty, whereas women will select partners

    who are good providers (Buss 1990, 1998).

    Beauty-status exchange accords with the

    popular conception of romantic partner selec-

    tion as a competitive market process, a con-

    ception widely accepted in both popular

    culture and academia. Under the market

    model of romantic relationship formation,

    individuals negotiate an informal exchange

    by trading their own assets for those of their

    partner. This market metaphor has been

    applied to the exchange of socioeconomic

    status for other purportedly desired resources,

    such as race (Fu 2001; Kalmijn 1993; Qian

    1997; Schoen and Wooldredge 1989; but see

    Rosenfeld 2005, 2010), homemaker skills(Becker 1991), youth (Coles and Francesconi

    2007; England and McClintock 2009), and

    physical attractiveness (Burdett and Coles

    2001; Carmalt et al. 2008; Elder 1969; Sassler

    and Joyner 2011; Stevens, Owens, and

    Schaefer 1990; Taylor and Glenn 1976; Udry

    1977). In popular culture, the concept of a

    gendered beauty-status exchange, in which an

    economically successful man partners with a

    beautiful trophy wife, is commonplace.In practice, the related exchange-based the-

    ory that individuals of low racial status but

    high socioeconomic status partner with those

    of high racial status and low socioeconomic

    status (status-caste exchange) also bolsters

    support for gender-stereotyped models of part-

    ner selection. Proponents argue that low-status

    white women partner with higher-status black

    men (Fu 2001; Kalmijn 1993; Qian 1997; Sch-

    oen and Wooldredge 1989) or that minoritywomen exchange beauty and sexual access for

    white mens income (Sassler and Joyner 2011).

    If this occurs, such couples would undermine

    racial boundaries but reinforce female eco-

    nomic dependency in marriage.

    These market-based exchange models are

    difficult to reconcile with the consistent empir-

    ical finding that romantic partners tend to

    match on many dimensions. In contradiction to

    beauty-status exchange theory, economically

    successful women partner with economically

    successful men (Sweeney and Cancian 2004),

    and physically attractive women partner with

    physically attractive men (Murstein and

    Christy 1976).1I attempt to resolve the empiri-

    cal paradox presented by the evidence of

    matching on physical attractiveness and on

    socioeconomic status with the conflicting lit-

    erature affirming beauty-status exchange. I

    argue that the positive individual-level correla-

    tion between physical attractiveness and socio-

    economic status, combined with a tendency for

    partners to match on both attractiveness and

    status, might be easily misconstrued

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    4/31

    McClintock 577

    as beauty-status exchange, particularly in

    regression models that do not fully account for

    matching.

    That said, although prior research may rely

    on faulty empirical models, it is possible thatbeauty-status exchange occurs in some cou-

    ples. Although they are competing forces,

    matching and exchange are not mutually

    exclusivesome couples might match while

    others engage in cross-trait exchange. Prior

    research proposes that certain types of cou-

    ples, such as less-committed or interracial

    couples (Sassler and Joyner 2011), are espe-

    cially likely to engage in beauty-status

    exchange. My analysis examines the preva-lence of matching and exchange among a

    representative sample of young couples and

    considers whether any subgroups dispropor-

    tionately engage in beauty-status exchange.

    Considering the contexts in which beauty-

    status exchange might predominate provides

    nuanced insight into processes of partner

    selection and highlights the differential mean-

    ings that couples may attach to their

    relationships.

    BACKGROUND

    Matching in Partner Selection

    The strongest force in partner selection is

    matching, or assortative mating (in marriage,

    endogamy): men and women select partners

    with characteristics similar to their own.

    Couples tend to be alike in education, age,

    race, and religion (Bereczkei and Csanaky

    1996; Blackwell and Lichter 2004; Mare

    1991; Rosenfeld 2008; Schoen and Cheng

    2006; Schoen and Weinick 1993; Stevens

    1991). Of most relevance to this study, there

    is strong evidence of matching on physical

    attractiveness (Berscheid et al. 1971; Carmalt

    et al. 2008; Feingold 1988; Kalick and Ham-

    ilton 1986; Murstein 1972; Murstein and

    Christy 1976; Stevens et al. 1990) and socio-

    economic status (Blackwell and Lichter 2004;

    Gardyn 2002; Mare 1991; Stevens 1991).

    Prior literature suggests at least three

    important forces generating matching: (1)

    homophily, that is, individuals preferences

    for similar (in-group) partners, (2) pressure

    from third parties to select similar partners,

    and (3) lack of contact with potential dissimi-

    lar (out-group) partners (Kalmijn 1998).These explanations are undoubtedly impor-

    tant and may suffice to explain matching on

    many traits, particularly traits for which pref-

    erences vary greatly between individuals. For

    these traits, such as religion, no one group is

    generally desired more than others. Instead,

    individuals want partners of their own group,

    whatever that may be. But for other traits,

    including income and physical attractiveness,

    more is often assumed to be better. For theseconsensually ranked traits, individuals may

    prefer partners who are superior to them-

    selves over partners who are their equals.

    Additionally, in the case of physical attrac-

    tiveness, it is not obvious that interested third

    parties would object to dissimilarity or that

    social-structural barriers prevent contact

    between more and less physically attractive

    individuals.

    A competitive marriage market modelmight explain some degree of matching on

    consensually ranked traits, including physical

    attractiveness. In a competitive market, every-

    one may desire the most beautiful and wealth-

    iest partners, but individuals will discover

    that the most desirable partner they can attract

    is one of their own level of desirability (see

    Burdett and Coles 1997, 1999, 2001; Choo

    and Siow 2006; Loughran 2002). Still, match-

    ing on specific traits is not the only possible

    outcome. Individuals might exchange a high

    level of one desirable trait for a high level of

    a different desirable trait in a partner, engag-

    ing in cross-trait exchange while matching on

    total desirability.

    I am interested in physical attractiveness

    and socioeconomic status because they are

    often thought to be consensually ranked traits,

    and existing theories of partner selection indi-

    cate that such traits might be either matched

    or exchanged. Despite the inherent tension

    between theories purporting beauty-status

    exchange and those arguing that couples will

    match on these traits, the theories are not

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    5/31

    578 American Sociological Review79(4)

    incompatible. Some couples might match

    while others exchange. Individuals might also

    seek approximate equality on both beauty and

    status but accept a limited tradeoff between

    the twosuch as a slightly less-attractivepartner with slightly higher status.

    Co-occurrence of Desirable Traits

    Analyses endorsing matching and those

    endorsing exchange generally overlook the

    co-occurrence of (un)desirable traitsthat

    beauty and status are positively correlated

    within individuals. Perhaps partially because

    physically attractive individuals are treatedpreferentially, they enjoy improved school

    performance, greater occupational success,

    and higher earnings (Clifford and Walster

    1973; Hamermesh and Biddle 1994; Haskins

    and Ransford 1999; Jackson, Hunter, and

    Hodge 1995; Langlois et al. 2000; Rosenblat

    2008; Rosenblat and Mobius 2006; Singer

    1964; Umberson and Hughes 1987; Wardle,

    Waller, and Jarvis 2002). Also, income may

    help individuals purchase goods and servicesthat enhance attractiveness, such as dental

    care and gym memberships. Some of the

    beauty-status correlation might be explained

    by rater bias: for example, individuals thought

    to be of higher-status nations are rated more

    favorably (Kowner 1996).2

    The co-occurrence of desirable traits

    encourages matching on multiple dimen-

    sionscouples who match on one trait tend

    toward similarity on related traits. This might

    lead to overestimating the strength of match-

    ing on any single dimension, considered alone

    (Kalmijn 1998). For example, if college grad-

    uates are (on average) better-looking than

    nongraduates, matching on college status

    facilitates matching on attractiveness. More

    subtly, because the within-individual co-

    occurrence of desirable traits might create a

    spurious between-partner cross-trait correla-

    tion of beauty and status, a bias toward

    observing mens status and womens beauty

    might cause matching to be misidentified as

    exchange. If couples match on college status

    and on attractiveness, college-educated men

    would have prettier wives than less-educated

    men (and college-educated women would

    have handsomer husbands). Because physi-

    cally attractive men and women average

    higher socioeconomic status, partner match-ing on status or attractiveness (or on both

    traits) would create a positive correlation

    between womens physical attractiveness and

    mens socioeconomic status, and between

    mens attractiveness and womens status,

    even in the absence of beauty-status exchange.

    In models that assume men value attrac-

    tiveness and women value status, couples

    would appear to engage in beauty-status

    exchange even when they match. For exam-ple, a high-earning man married to a pretty

    wife might be interpreted as beauty-status

    exchange by a researcher who only observes

    male status and female beautybut if the

    man is handsome and his wife a high-earner,

    then the couple is matched on both traits.

    Most prior studies of beauty-status exchange

    assume a gendered importance of beauty and

    status, overlooking mens attractiveness and

    womens status and potentially misidentify-ing matching as exchange. Analyses assum-

    ing a stereotypically gendered importance of

    beauty and status could thus erroneously, and

    unintentionally, perpetuate these very stereo-

    types. Also, because most prior studies

    assume gendered exchange, it is unclear

    whether patterns of beauty-status exchange

    (insofar as they exist after controlling for

    matching) are gender-symmetric.

    Evidence of Exchange

    Despite strong evidence of matching on phys-

    ical attractiveness (Carmalt et al. 2008; Fein-

    gold 1988; Murstein 1972; Murstein and

    Christy 1976; Stevens et al. 1990) and on

    socioeconomic status (Blackwell and Lichter

    2004; Gardyn 2002; Mare 1991; Stevens et al.

    1990), prior studies argue that physical attrac-

    tiveness is traded for socioeconomic status

    (Bjerk 2009; Carmalt et al. 2008; Elder 1969;

    Taylor and Glenn 1976; Udry 1977). Specifi-

    cally, early studies of beauty-status exchange

    in partner selection propose a gendered

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    6/31

    McClintock 579

    exchange of womensbeauty for mensstatus.

    Such studies find that both physical attrac-

    tiveness and education help a woman achieve

    upward mobility through marriage (defined

    as marrying a man of higher occupationalstatus than her father [Elder 1969; Udry

    1977]) and help her marry a man of high

    occupational status, in absolute terms (Taylor

    and Glenn 1976). However, these analyses

    exclude mens physical attractiveness and

    therefore do not address how the co-occur-

    rence of desirable traits within individuals

    might encourage matching and might also

    generate the illusion of exchange.3 Control-

    ling for bothpartners physical attractivenessmay not eliminate the relationship between

    female beauty and male status, but it should at

    least reduce this relationship substantially.

    A recent article using the National Longi-

    tudinal Study of Adolescent Health (Add

    Health) Romantic Pair data, a nationally rep-

    resentative sample of young adult dating,

    cohabiting, and marital partners, finds that for

    both genders, personal assets including edu-

    cation predict having a physically attractivepartner (Carmalt et al. 2008). The authors

    interpret this as evidence of cross-trait beauty-

    education exchange, but they fail to control

    for the partners education. Because educa-

    tion is strongly associated with physical

    attractiveness within individuals (beauty and

    education co-occur), it is unclear from this

    study whether education is exchanged to

    attain an attractive partner or whether the

    findings are spurious and result from match-

    ing. Therefore, I use the same data to evaluate

    patterns of beauty-status exchange while con-

    trolling for matching.

    Some individuals certainly intend to trade

    high levels of one trait for high levels of a dif-

    ferent trait. Personal and online dating adver-

    tisements are often couched in deal-making

    language (Alterovitz and Mendelsohn 2009;

    Davis 1990; Harrison and Saeed 1977). These

    advertisements imply that some women hope

    to trade their physical attractiveness for mens

    financial security, and some men hope to trade

    their socioeconomic status for a physically

    attractive but lower-status partner. But

    experimental studies indicate that women (and

    men) will not compromise on physical attrac-

    tiveness (Li et al. 2002; Li and Kenrick 2006;

    Sprecher 1989). In practice, individuals might

    be unwilling to make the cross-trait exchangesthat they anticipated making. Moreover,

    women and men pursuing a gendered beauty-

    status exchange in their advertisements might

    be overlooking the importance of class-based

    cultural compatibility. They may ultimately

    value cultural similarity more than similarity

    on tangible measures of status (income or

    degree attainment) (Kalmijn 1994), but they

    may realize the cultural importance of socio-

    economic status only after meeting status-disparate dates. For example, a man might

    attach little importance to a prospective wifes

    income or educational attainment per se, but

    he might discover that the women with whom

    he is most interpersonally compatible are

    women of his own socioeconomic status.

    Is More of a Good Thing AlwaysBetter?

    Prior studies espousing beauty-status

    exchange implicitly assume that physical

    attractiveness and social status are consensu-

    ally ranked traits: that is, these traits can be

    measured and partners with higher levels are

    generally more desired (e.g., Elder 1969;

    Taylor and Glenn 1976). Socioeconomic sta-

    tus is an abstract concept, but it is often

    approximated by education, income, or occu-

    pation-tangible and quantifiable characteris-

    tics. Despite some inevitable subjectivity in

    evaluating physical attractiveness, prior stud-

    ies report high inter-rater consensus (Langlois

    et al. 2000; Murstein 1972). In the United

    States, there may be racial differences in ideal

    body shape (Cohn and Adler 1992; Lovejoy

    2001; Webb, Looby, and Fults-McMurtery

    2004), but assessment of facial attractiveness

    does not vary by race (Cunningham et al.

    1995; Moss, Miller, and Page 1975). Attrac-

    tiveness and status are thus quantifiable.

    However, neither men nor women report

    attaching much importance to either trait. For

    example, when asked to rank the importance

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    7/31

    580 American Sociological Review79(4)

    of 76 traits in a romantic partner, no measures

    of appearance or socioeconomic status made

    the top 10 (Buss and Barnes 1986). This and

    similar studies find that men value appear-

    ance more than women, whereas womenvalue socioeconomic indicators more than

    men, but neither gender ranks physical attrac-

    tiveness or socioeconomic status highly

    (Furnham 2009; Howard, Blumstein, and

    Schwartz 1987; Nevid 1984). (This gender

    difference may be declining in recent cohorts

    [Buss et al. 2001; Regan and Joshi 2003].)

    In contrast to self-reported preferences,

    experimental studies find that physical attrac-

    tiveness is highly valued by both genders, andthat women also value mens socioeconomic

    status (Li et al. 2002; Li and Kenrick 2006;

    Sprecher 1989). Similarly, recent speed-

    dating studies find that physical attractiveness

    and earnings potential are both strong predic-

    tors of attraction (Eastwick and Finkel 2008;

    Fisman et al. 2006; Luo and Zhang 2009).

    Consistent with these findings, women and

    men using an online-dating website valued

    physical attractiveness highly (Hitsch, Hort-acsu, and Ariely 2010). Evidence from speed-

    dating and online-dating studies is mixed

    regarding gender differences in the relative

    importance of attractiveness and earnings

    potential.

    Arguably, acted preferences in experi-

    ments and real dating situations may be more

    genuine indicators of true preferences than

    are stated preferences (McClintock 2011)

    and acted preferences indicate that when it

    comes to beauty and status in romantic part-

    ners, more is better. This is a necessary but

    not sufficient condition for exchange to occur.

    Exchange requires not only that these traits

    are valued, but also that they are substituta-

    ble. Yet individuals might be unwilling to

    compromise on one dimension (e.g., by

    accepting a homely partner), even when com-

    pensated on another dimension (if the homely

    partner were high-status). Indeed, acted pref-

    erences suggest that both genders consider

    physical attractiveness a necessity and disre-

    gard potential partners who fail their attrac-

    tiveness criteria (see Li and Kenrick 2006).

    Union Status and Duration

    Most studies purporting beauty-status

    exchange restrict their analysis to married

    couples. Only the most recent of these articles(Carmalt et al. 2008) uses dating, cohabiting,

    and married couples. I use these same data,

    the National Longitudinal Study of Adoles-

    cent Health Romantic Pair sample (2001 to

    2002;N= 1,507). Insofar as commitment can

    be assumed to increase as couples move from

    dating to cohabitation to marriage, couples in

    more committed relationships are more simi-

    lar in race, education, and other traits (Black-

    well and Lichter 2004; Joyner and Kao 2005).Individuals may hold looser partner criteria

    for short-term relationships, and dissimilar

    couples may face higher dissolution rates

    (McClintock 2010; Wang, Kao, and Joyner

    2006). If beauty-status exchange is a form of

    dissimilarity that increases instability, it would

    be most prevalent in dating couples, less

    prevalent among cohabiters, and least preva-

    lent among married couples. Likewise,

    beauty-status exchange would be less com-mon in relationships of longer duration. In

    addition to winnowing (the selective disso-

    lution of less similar couples), romantic part-

    ners may grow more similar over time due to

    shared lifestyle (e.g., eating habits, exercise,

    or encouragement to earn a higher degree).

    However, entering into a beauty-status

    exchange is most rational if one anticipates

    marriage, because it is in marriage that eco-

    nomic resources are most often shared. If so,

    exchange would be most prevalent among

    married couples, because not all dating and

    cohabiting couples anticipate marriage. Cohab-

    iting couples would be especially unlikely to

    exchange beauty and status, because they are

    generally more egalitarian than married cou-

    ples (Brines and Joyner 1999) and are less

    likely to pool resources (Hamplova and Le

    Bourdais 2009). Indeed, some research sug-

    gests that patterns of exchange evident among

    married couples are not viable among cohabit-

    ers, including the exchange of housework for

    economic support (Brines and Joyner 1999)

    and the exchange of racial caste for

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    8/31

    McClintock 581

    socioeconomic status (Blackwell and Lichter

    2000; but see Rosenfeld 2005). If my analysis

    produces evidence of beauty-status exchange,

    differences in its prevalence by union status

    (dating, cohabiting, married) might clarifywhether it is a source of instability or a viable

    long-term strategy.

    Interracial Couples

    In a classic variant of cross-trait exchange in

    partner selection, individuals of low racial

    status but high socioeconomic status partner

    with those of high racial status but low socio-

    economic status (Davis 1941; Fu 2001;Kalmijn 1993; Merton 1941; Schoen and

    Wooldredge 1989). In practice, race-status

    (status-caste) exchange, like beauty-status

    exchange, is gender-stereotypical: lower-

    status white women are thought to partner with

    higher-status black men (Fu 2001; Kalmijn

    1993; Qian 1997; Schoen and Wooldredge

    1989). Although not entirely analogous to

    beauty-status exchange (individuals desire

    partners of higher status and greater beauty,but generally prefer partners of their own race

    [Yancey 2009]), race-status exchange theory

    suggests that interracial couples might be

    uniquely prone to engage in cross-trait

    exchange. Indeed, Sassler and Joyner (2011)

    find that minority women exchange beauty

    and sexual access for white mens social sta-

    tus (but their test is incomplete because they

    omit mensphysical attractiveness).

    However, just as beauty-status exchange

    conflicts with couple matching on beauty and

    status, race-status exchange conflicts with

    matching on race and status. Empirical sup-

    port for race-status exchange is mixed and

    controversial (Gullickson and Fu 2010;

    Kalmijn 2010; Rosenfeld 2005, 2010). It is

    beyond the scope of this article to test whether

    couples engage in race-status exchange. I do,

    however, consider whether patterns of beauty-

    status exchange might differ for interracial

    couples. Part A of the online supplement

    (http://asr.sagepub.com/supplemental)

    addresses beauty-status exchange among

    interracial couples in greater depth.

    HYPOTHESES

    Prior research demonstrates some support for

    the theory that individuals trade one desirable

    trait to obtain a partner with more of a differ-ent desirable trait, but in the case of beauty

    and status there is stronger evidence in sup-

    port of matching. I expect that patterns of

    beauty-status exchange found in earlier stud-

    ies are, at least in part, artifacts of matching

    on partner traits not controlled for in the

    model.However, prior research proposes that

    the prevalence of beauty-status exchange may

    differ for certain subgroups, such as interra-

    cial and less-committed couples.Understanding the prevalence of beauty-

    status exchange is important for theoretical

    conceptions of partner selection and for ste-

    reotypes regarding the importance of beauty

    and status for women and men. Gendered

    beauty-status exchange is the most prominent

    example of cross-trait exchange in the part-

    ner market, especially in popular culture.

    Questioning the frequency of cross-trait

    exchange in a context in which it is generallyassumed to occur does not challenge the mar-

    ket model in all contexts, but it weakens its

    status as thepredominantparadigm for under-

    standing partner selection processes. Addi-

    tionally, if there is no evidence of gendered

    beauty-status exchange, this challenges the

    evolutionary model of partner selection and

    discredits common assumptions about wom-

    ens and mens priorities for partners. Finally,

    understanding the social contexts in which

    beauty-status exchange may occur provides

    insight into the varying meanings of young-

    adult romantic relationships.

    DATA AND MEASURES

    Data

    I use data from the 2001 to 2002 National

    Longitudinal Study of Adolescent Health

    Romantic Pair sample, a supplementary data-

    set to the 1994 to 2008 National Longitudinal

    Study of Adolescent Health (Add Health).

    Add Health is a nationally representative,

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    9/31

    582 American Sociological Review79(4)

    longitudinal survey of adolescents in grades 7

    through 12 at the initial interview (Chantala

    2006; Harris et al. 2009). In the third wave of

    data collection (Wave III), respondents, who

    were mainly in their early 20s, providedinformation on romantic relationships. From

    current relationships of minimum three-

    months duration, approximately 500 dating,

    500 cohabiting, and 500 marriage partners

    were selected to complete a slightly modified

    version of the Wave III interview. These

    1,507 couples (3,014 individuals) make up

    the Romantic Pair sample.

    Women in the Romantic Pair sample aver-

    age about 22 years old and men average about23.5 years. Except for Carmalt and colleagues

    (2008) who use these same data, prior analyses

    use small convenience samples (Elder 1969;

    Stevens et al. 1990) or datasets that measure

    only one partners attractiveness (Elder 1969;

    Taylor and Glenn 1976; Udry 1977). Respond-

    ents in the Romantic Pair data are equivalent in

    age to at least two of the samples used previ-

    ouslyCarmalt and colleagues (2008) and

    Stevens and colleagues (1990). The samplesused by Udry (1977) and Taylor and Glenn

    (1976) benefit from substantial age variation

    (25 to 40 years), but they measure only one

    partners physical attractiveness, and attrac-

    tiveness ratings were often made several years

    after marriage. Thus, although the Romantic

    Pair data may not be ideal due to the samples

    young average age and limited age-range (18

    to 43 years), it improves upon datasets used in

    prior analyses. Like prior analyses, this study

    inevitably excludes single individuals; supple-

    mental analyses suggest that attractiveness and

    status promote partnering (see Part B in the

    online supplement).

    Measures

    Add Health interviewers rated respondents

    physical attractiveness using five levels: (1)

    very unattractive to (5) very attractive. An

    average of several observers ratings is ideal,

    but prior studies indicate high inter-rater reli-

    ability in evaluating attractiveness (Langlois

    et al. 2000; Murstein 1972). In the Add Health

    data, interviewers ratings are positively and

    significantly correlated across waves: the cor-

    relation between Wave I and Wave III ratings

    (six to seven years apart) is .17 (p < .001);

    between Wave II and Wave III ratings (fiveyears apart), the correlation is .20 (p< .001).

    When the same interviewer rated both part-

    ners, there is no evidence of unconsciously

    exaggerating their similarity. Indeed, the cor-

    relation of her and his physical attractiveness is

    slightly lower when both partners were inter-

    viewed by the same interviewer (not shown).

    Interviewers also rated respondents

    grooming and personality attractiveness on

    five-level scales, yielding a three-item per-sonal attractiveness index (physical, groom-

    ing, personality). I primarily present models

    using the single-item measure of physical

    attractiveness, but results using the index are

    similar. Body mass index (BMI) is a ratio of

    height to weight approximating thinness or

    overweight. Self-rated healthranges from (1)

    fair or poor to (4) excellent.

    Years of educationrefers to years of com-

    pleted education. Among respondents, 17 per-cent of men and 25 percent of women were

    enrolled in college full-time, but only in 11

    percent of couples were both partners full-

    time students. Expected/completed college

    graduation status is a dichotomous variable

    indicating whether a respondent was a college

    graduate or enrolled in a four-year degree

    program versus having never attended college

    or having dropped out. This variable addresses

    right-censoring.

    To measuresocioeconomic status,I use the

    Duncan Socioeconomic Index (SEI; Ruggles

    et al. 2010), a measure of occupational pres-

    tige that I obtained from U.S. Census data

    from the Integrated Public Use Microdata

    Series (IPUMS; Minnesota Population Center

    2013). I use both current and forecast (future)

    SEI. To create forecast SEI, I regressed Wave

    IV SEI (five years later; available only for

    original respondents) on race, age, educational

    attainment, current SEI, picture-vocabulary

    test score at Wave III, parental income, paren-

    tal education, and parental SEI. I estimated

    models separately by gender. I then used

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    10/31

  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    11/31

    584 American Sociological Review79(4)

    Log-linear and negative binomial regres-

    sion models are also particularly appropriate

    for identifying patterns of matching or

    exchange. Controlling for the distribution of

    beauty and status among partnered men andwomen, these models test which pairings are

    overrepresented in the data (e.g., whether

    beautiful low-status women are disproportion-

    ately paired with unattractive high-status

    men). Parameters directly model the tendency

    for partners to match (by having equal levels

    of a given trait) or exchange (when one part-

    ner has a higher level of one trait and the other

    partner has a higher level of another trait).

    The Current Analysis

    I begin by describing the co-occurrence of

    desirable traits at the individual and couple

    level to illustrate the importance of including

    complete information on both partners char-

    acteristics in regression models testing

    exchange theories. To search for evidence of

    exchange, I examine the differences between

    partners endowments of desirable traits.Finally, because it is couples who differ on a

    given trait that might engage in cross-trait

    exchange, I look for patterns of exchange

    among this subgroup of couples.

    I next consider multivariate regression

    models, estimating models that ignore match-

    ing and the within-individual correlation of

    desirable traits and then adding variables to

    fully control for these forces. Consistent with

    prior work, I estimate models separately by

    gender. All models include age, race, preg-

    nancy status, relationship duration, and union

    status. Next, to test for exchange more

    directly, I estimate regression models using

    difference measures (his minus her level of

    each trait). Finally, I estimate negative bino-

    mial regression models, a generalization of

    log-linear models that are preferred when

    data are scarce and over-dispersion is prob-

    lematic (King 1989; Long 1997; Long and

    Freese 2006). In these data, relatively few

    couples have disparate levels of attractiveness

    and socioeconomic statusdata are scarce in

    these cells. I use various measures and model

    specifications to test the robustness of results.

    I primarily present results using the entire

    sample, but some key results are also pre-

    sented for subgroups suspected of engaging

    in beauty-status exchangecouples who dif-

    fer on attractiveness (N = 810) or collegestatus (N= 242), minority-female white-male

    couples (N = 95), and couples involving a

    working-class-origin woman (N= 1,146) or a

    five-year or larger age gap favoring the man

    (N = 206). I also discuss models estimated

    separately by union status; models that inter-

    act the effects of interest with union status,

    relationship duration, class background, and

    race; and models restricted to couples in the

    top 25th percentile of the age distribution(women 23 to 40 years and men 24 to 43

    years, N = 378), non-student couples (N =

    1,341), couples interviewed by the same

    interviewer (N= 1,305), and minority-white

    couples (N= 209).

    Although no one variable is missing for a

    large number of cases, many cases are miss-

    ing data on at least one variable. Dropping

    cases with missing data would reduce the

    sample size to 1,408 and would likely intro-duce bias (Acock 2005). I use the ICE (Impu-

    tation by Chained Equations) procedure in the

    statistical software program Stata 10.1

    (Royston 2004) to impute missing data using

    switching regression, an iterative multivaria-

    ble regression technique. I show results from

    regression models estimated using the MIM

    procedure in Stata 10.1 (Royston 2004).

    Descriptive statistics and regression models

    calculated by dropping cases with missing

    data are similar to those using MIM (not

    shown).

    RESULTS

    Sample Characteristics

    Table 1 presents summary statistics on all

    variables, by gender. Women are rated more

    physically attractive than men. This gender

    difference in physical attractiveness exists in

    the full Add Health sample (including among

    single individuals; see Table S3 in the online

    supplement), perhaps because women invest

    more in personal care and cosmetics. If women

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    12/31

    McClintock 585

    in prior cohorts were rated as more attractive

    than men, this may have created the illusion of

    exchange. In prior cohorts, many women mar-

    ried higher-status men simply because, on

    average, men had more education and higher

    occupational status than did women. Even if

    all couples sought equality in physical attrac-

    tiveness and status, men would have had

    higher status while women would have been

    more attractive. But this would have resulted

    from gender differences in average endow-

    ments of attractiveness and status rather than

    from a beauty-for-status exchange. In con-

    trast, in these data women surpass men in

    educational attainment and occupational sta-

    tus (SEI). This gender reversal reflects changes

    in the general populationwomen now earn

    more college degrees than men (DiPrete and

    Table 1. Mean Characteristics by Gender; National Longitudinal Study of Adolescent HealthRomantic Pair Sample (2001 to 2002); N= 1,507

    Women Men

    Mean orProportion SD Mean orProportion SD

    Attractiveness

    Physical attractivenessa 3.63*** .86 3.44 .75

    Personal attractiveness indexa 3.67*** .69 3.48 .62

    Body mass index (BMI) 26.49*** 6.67 27.15 5.64

    Self-rated healthb 2.92*** .87 3.10 .84

    Socioeconomic Status

    Expected college graduatec .27*** .22

    Years of completed education 12.96** 1.96 12.74 1.98

    Current Duncan socioeconomic index (SEI)d 28.66 27.34 28.31 24.89

    Forecast Duncan socioeconomic index (SEI)d 46.94*** 8.74 39.69 11.73 Current income ($10,000s) 11.96*** 17.69 19.69 22.65

    Forecast income ($10,000s) 29.46*** 12.91 46.83 13.09

    Working-class social originse .76 .74

    Demographic Characteristics

    Age (years) 21.85*** 2.37 23.48 3.30

    Race

    White .59 .59

    Black .17 .18

    Other .24 .23

    Pregnant .07

    Relationship CharacteristicsRelationship duration (months) 38.66 26.66 38.29 26.76

    Married .36 .36

    Cohabiting .36 .36

    Dating .28 .28

    Note:Descriptive statistics use multiple imputed datasets to estimate missing values.aScored from 1 (very physically unattractive) to 5 (very physically attractive). The personalattractiveness index reflects respondents average rating on three measures of attractiveness: physical,personality, and grooming.

    bFrom (1) fair or poor to (4) excellent.cThis dichotomous variable indicates respondents who have completed a four-year college degree or are

    currently enrolled in a four-year degree program.dA measure of occupational prestige. Ratings range from 0 (low) to 96 (high). Forecast SEI is projectedSEI at Wave IV when respondents will be in their late 20s.eI define individuals of working-class social origins as having a father who worked in an occupationwith a Hollingshead rank of four or lower.*p< .05; **p< .01; ***p< .001;p-values indicate significant gender differences (two-tailed tests).

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    13/31

    586 American Sociological Review79(4)

    Table 2. Within-Individual Correlation of Desirable Characteristics; National LongitudinalStudy of Adolescent Health Romantic Pair Sample (2001 to 2002); N= 1,507

    Women Men

    Correlation of HerAttractiveness and

    Her Status

    Correlation of HisAttractivenessand His Status

    Attractiveness EDUd GRADe SEIf INCg EDUd GRADe SEIf INCg

    Physical attractiveness .127***a .155***a .175***a .065*a .166***a .156***a .132***a .096***a

    Personal attractivenessindex

    .179***b .186***b .222***b .104***b .230***b .179***b .197***b .120***b

    BMI, reverse-codedc .140***b .207***b .207***b .043b .015b .071**b .011b .028b

    Self-rated health .200***a .191***a .167***a .102***a .172***a .166***a .124*a .102***a

    Note:Descriptive statistics use multiple imputed datasets to estimate missing values.aCorrelations and significance tests involving one or more ordinal variables are calculated usingSpearmans Rho.

    bCorrelations and significance tests involving only interval variables are calculated using Pearsonscorrelation.cBody mass index. Reverse-coded so that higher values indicate thinner (normatively more desirable)physiques.dEDU is years of completed education.eGRAD is expected college graduation status.fSEI is projected future occupational status.gINC is projected future income.*p< .05; **p< .01; ***p< .001 (two-tailed tests).

    Buchman 2006), and single, childless, young

    women out-earn their male counterparts

    (Wiseman 2010). These gender differences in

    education and occupational status observed in

    the Add Health data are also evident among

    respondents of a similar age in the 2000 U.S.

    Census (authors calculations; not shown).

    Co-occurrence of Desirable Traits

    Table 2 demonstrates that the various dimen-

    sions of desirability are strongly and posi-

    tively correlated within individuals. Because

    physically attractive individuals tend to be

    high in socioeconomic status (and vice versa),

    partner matching on one or both of these traits

    might generate the illusion of cross-trait

    between-partner exchange. The within-indi-

    vidual correlations between desirable traits are

    not strong enough to preclude exchange, butthey are strong enough to generate spurious

    evidence of exchange. Additionally, when

    desirable traits co-occur, matching on any one

    trait encourages matching on related traits.

    Between-Partner Similarity

    Table 3 displays strong evidence of matching on

    physical attractiveness, education, and occupa-

    tional status (SEI). The correlations between her

    and his expected college graduation status

    (.575), years of completed education (.557), and

    SEI (.546) are especially strong. The between-

    partner correlation of attractiveness (.256) is

    similar in magnitude to equivalent correlationsin all other samples known to the author (Barelds

    et al. 2011; Stevens et al. 1990). None of the

    statistically significant between-partner within-

    trait correlations are negative, as predicted by an

    exchange model in which high-status but

    homely individuals are paired with low-status

    but good-looking partners.

    The various measures of one partners

    desirability are also positively and signifi-

    cantly correlated with other measures of theother partners desirability. For example, the

    correlation between her physical attractive-

    ness and his years of completed education is

    .186 (p< .001). Prior studies present similar

    at SANTA CLARA UNIV on August 9, 2014asr.sagepub.comDownloaded from

    http://asr.sagepub.com/http://asr.sagepub.com/http://asr.sagepub.com/
  • 8/12/2019 American Sociological Review 2014 McClintock 575 604

    14/31

    587

    Table3.

    Betw

    een-PartnerCorrelationofW

    omensandMensCharacteristics;NationalLongitudinalS

    tudyofAdolescentHealthRo

    manticPair

    Sample(200

    1to2002);N=1,5

    07

    MalePartnersCharacteristics

    Attractiveness

    SocioeconomicStatus

    FemalePartnersCharacteristics

    ATT

    RACT

    PER

    BM

    I

    HEALTH

    EDU

    GRAD

    SEI

    INC

    Attractiveness

    Physicalattractiveness(ATTRACT)

    .256***a

    .297***a

    .070**a

    .099***a

    .186***a

    .142***a

    .130***a

    .057a

    Personalatt

    ractivenessindex(PER)

    .306***a

    .386***b

    .050b

    .123***a

    .190***b

    .145***b

    .165***b

    .087**b

    BMI,reverse-codedc

    .134***a

    .134***b

    .266***b

    .138***a

    .182***b

    .191***b

    .164***b

    .112***b

    Self-ratedhealth(HEALTH)

    .104***a

    .133***a

    .044a

    .172***a

    .182***a

    .157***a

    .145***a

    .111***a

    SocioeconomicStatus

    Yearsofcom

    pletededucation(EDU)

    .143***a

    .188***b

    .0

    06b

    .140***a

    .557***b

    .456***b

    .450***b

    .308***b

    Expectedco

    llegegraduate(GRAD)

    .119***a

    .150***b

    .058*b

    .147***a

    .499***b

    .575***b

    .481***b

    .295***b

    Forecastsocioeconomicindex(SEI)

    .141***a

    .176***b

    .048b

    .099***a

    .525***b

    .555***b

    .546***b

    .351***b

    Forecastinc

    ome(INC)

    .080**a

    .108***b

    .036b

    .066*a

    .358***b

    .333***b

    .373***b

    .330***b

    Note:Descriptivestatisticsusemultipleimputeddatasetstoestimatemissingva

    lues.

    aCorrelationsandsignificancetestsinvolvingoneormoreordinalvariablesarec

    alculatedusingSpearmansRho.

    bCorrelationsandsignificancetestsinvolvingonlyintervalvariablesarecalculatedusingPearsonscorrelation.

    cBodymassin

    dex.

    Reverse-codedsothathighe

    rvaluesindicatethinner(normat

    ivelymoredesirable)physiques.

    *p