《性别、亲密关系与社会》课程教学资源：Trends and Variation in Assortative Mating：Causes and Consequences

团购合买资源类别：文库，文档格式：PDF，文档页数：23，文件大小：214.75KB

INTRODUCTION degree to which populations intermarry tells This article reviews recent studies of trends us about the openness of social boundaries. and variation in assortative mating,the nonran- Studies of assortative mating are also of interest dom matching of individuals into relationships. for personal reasons.People care deeply about People match assortatively in many ways,such who their romantic partners are.Matching as into jobs or friendships;this review focuses patterns are far from random,and the charac- on matching in romantic relationships.I follow teristics of partners affect relationship quality, convention and use the term assortative mating, satisfaction,and stability. borrowed from biological studies of mating in This article builds on previous reviews by Kalmijn (1998)and Blossfeld(2009).Kalmijn's animal populations,but my focus is on roman- tic relationships broadly construed rather than comprehensive review covered socioeconomic, on the characteristics of couples engaged in sex racial/ethnic,and religious matching studies and reproduction. written primarily in the 1980s and 1990s. Studies of assortative mating span many Blossfeld's review was more targeted,focusing disciplines.Sociologists interested in assorta- on trends and variation in educational assorta- tive mating typically focus on matching across tive mating.Like Kalmijn,I review socioeco- salient social boundaries,most often socio- nomic,racial/ethnic,and religious matching, economic status,race/ethnicity,and religion. but I update his review with studies published There is also a vibrant literature in psychology, since the late 1990s.Each of these topics could distinguished by a focus on matching on per- fill its own article.I include all three because sonality characteristics,although these studies even though the theories,concepts,and issues also often include demographic traits.Interest overlap,connections and insights from one in assortative mating in economics,which tends area are rarely applied to the others.Unlike to focus on spouses'economic resemblance,has Kalmijn's review,this article is primarily lim- grown considerably over the past 20 years.This ited to studies that inform our understanding of review focuses on the three types of assorta- trends and variation in assortative mating.This tive mating most common in the sociology and focus necessarily omits some active areas in the economics literature:(a)socioeconomic sta- literature,such as point-in-time studies of in- tus (including education,occupation,income, dividuals'preferences for mates and patterns of and class background);(b)race/ethnicity;and matching within a single context.Unlike both (c)religion. Kalmijn and Blossfeld,I devote considerable at- One of the major reasons why assortative tention to the consequences of assortative mat- mating has received so much attention in the ing.Research on this topic has grown substan- biological and social sciences is that it fun- tially,but findings from these studies have not damentally shapes the characteristics of yet been the subject of a comprehensive review. populations.It organizes people into families and determines the characteristics of parents. HOW ARE PATTERNS OF Because spouses (and to a lesser extent cohabit- ASSORTATIVE MATING ing couples)share resources,assortative mating GENERATED? affects individuals'access to resources and the The conceptual framework guiding most distribution of resources across families. studies of assortative mating is the notion of Moreover,to the extent that parents influence the marriage market.Individuals search for their children's characteristics,assortative partners in a market in which people have mating shapes the characteristics of future preferences for mates but face constrained op- populations.Boundaries between social groups portunities(Becker 1974,Goode 1970 [1963], are maintained through assortative mating and Mortensen 1988,Oppenheimer 1988).Search weakened through intermarriage.Thus,the is costly (in terms of both time and money),and 452 Sdrwartz

SO39CH22-Schwartz ARI 26 June 2013 15:20 INTRODUCTION This article reviews recent studies of trends and variation in assortative mating, the nonrandom matching of individuals into relationships. People match assortatively in many ways, such as into jobs or friendships; this review focuses on matching in romantic relationships. I follow convention and use the term assortative mating, borrowed from biological studies of mating in animal populations, but my focus is on romantic relationships broadly construed rather than on the characteristics of couples engaged in sex and reproduction. Studies of assortative mating span many disciplines. Sociologists interested in assortative mating typically focus on matching across salient social boundaries, most often socioeconomic status, race/ethnicity, and religion. There is also a vibrant literature in psychology, distinguished by a focus on matching on personality characteristics, although these studies also often include demographic traits. Interest in assortative mating in economics, which tends to focus on spouses’ economic resemblance, has grown considerably over the past 20 years. This review focuses on the three types of assortative mating most common in the sociology and economics literature: (a) socioeconomic status (including education, occupation, income, and class background); (b) race/ethnicity; and (c) religion. One of the major reasons why assortative mating has received so much attention in the biological and social sciences is that it fundamentally shapes the characteristics of populations. It organizes people into families and determines the characteristics of parents. Because spouses (and to a lesser extent cohabiting couples) share resources, assortative mating affects individuals’ access to resources and the distribution of resources across families. Moreover, to the extent that parents influence their children’s characteristics, assortative mating shapes the characteristics of future populations. Boundaries between social groups are maintained through assortative mating and weakened through intermarriage. Thus, the degree to which populations intermarry tells us about the openness of social boundaries. Studies of assortative mating are also of interest for personal reasons. People care deeply about who their romantic partners are. Matching patterns are far from random, and the characteristics of partners affect relationship quality, satisfaction, and stability. This article builds on previous reviews by Kalmijn (1998) and Blossfeld (2009). Kalmijn’s comprehensive review covered socioeconomic, racial/ethnic, and religious matching studies written primarily in the 1980s and 1990s. Blossfeld’s review was more targeted, focusing on trends and variation in educational assortative mating. Like Kalmijn, I review socioeconomic, racial/ethnic, and religious matching, but I update his review with studies published since the late 1990s. Each of these topics could fill its own article. I include all three because even though the theories, concepts, and issues overlap, connections and insights from one area are rarely applied to the others. Unlike Kalmijn’s review, this article is primarily limited to studies that inform our understanding of trends and variation in assortative mating. This focus necessarily omits some active areas in the literature, such as point-in-time studies of individuals’ preferences for mates and patterns of matching within a single context. Unlike both Kalmijn and Blossfeld, I devote considerable attention to the consequences of assortative mating. Research on this topic has grown substantially, but findings from these studies have not yet been the subject of a comprehensive review. HOW ARE PATTERNS OF ASSORTATIVE MATING GENERATED? The conceptual framework guiding most studies of assortative mating is the notion of the marriage market. Individuals search for partners in a market in which people have preferences for mates but face constrained opportunities (Becker 1974, Goode 1970 [1963], Mortensen 1988, Oppenheimer 1988). Search is costly (in terms of both time and money), and 452 Schwartz Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

people compete with one another for preferred wants to partner“down,”and everyone ends up mates.The search for a romantic partner has with someone roughly similar to themselves.A often been compared with the search for a job. small literature attempts to adjudicate between Just as people have a reservation wage below these two hypotheses (Dimaggio Mohr which they will not take a job,they may also 1985,Kalmijn 1994),with recent studies in have a set of minimum qualifications they seek economics using novel data from internet in a partner below which they will not form dating websites,speed dating experiments,and a match.People may update their preferences marriages among movie stars (Bruze 2011, on the basis of their experience in the marriage Fisman et al.2006,Hitsch et al.2010).Overall, market,and the complex interaction between the evidence favors the matching hypothesis, people's potentially shifting preferences and but whether matching or competition prevails their constraints generates observed matching may vary by the trait in question,e.g.,compe- patterns.These ideas underlie many behavioral tition for income and matching on education, models of mate selection,although various race/ethnicity,and religion(Hitsch etal.2010). simplifying assumptions are made(e.g.,Burdett All these studies were conducted using data Coles 1997,Logan et al.2008,Wong 2003). from the United States,however,and whether They have primarily been applied to marriage matching or competition characterizes the but also apply to the search for other types of sorting process may vary by context. romantic partners (e.g,Choo Siow 2006, Becker predicted that negative assorta- Sahib Gu 2002). tive mating would occur for traits that are Becker's (1974,1981)economic model of substitutes.Traits are substitutes when there marriage is often evoked in studies of assor-are gains to specialization.Becker reasoned 宝 tative mating because of its clear predictions that,just as factory output can increase when about matching patterns.He predicted that like individuals specialize rather than complete all will marry like when traits are complements: parts of a process,the gains to marriage can be characteristics such as lifestyle,attractive-greater when members of a couple specialize, ness,and religion.Becker referred to this and individuals would thus sortinto marriageto phenomenon as positive assortative mating.maximize these gains.Although specialization The terms endogamy and exogamy are also could occur on numerous household tasks(e.g., often used,respectively,for in-or out-group I'm better attidying up,and my partner is better marriage (e.g.,by religion,race/ethnicity,or at washing dishes),Becker primarily focused on nativity),and homogamy and heterogamy specialization in the broad categories of market for marriages between people with similar or work versus household work.As others have Aold dissimilar traits (e.g.,by years of schooling,pointed out,Becker's ideas are reminiscent income,or attractiveness). of Parsons's (1949)classic discussion of the Other scholars have noted that homogamy conjugal family,in which the division of labor can be generated by two combinations of between husbands and wives serves to reduce preferences for mates:(a)Both sides of the competition between the sexes and promote market prefer partners with characteristics marital stability.The gains to specialization led similar to their own(the matching hypothesis),Becker to predict negative assortative mating or(b)both sides of the market prefer partners on wages such that high-wage men(or women) with more of a characteristic (the competition marry low-wage women (or men). hypothesis).The matching hypothesis leads to Exchange theory (Davis 1941,Merton homogamy because everyone is searching for 1941)also predicts negative assortative mating, someone similar to themselves.The competi-but for different reasons.In the absence of tion hypothesis leads to homogamy because,homogamy,it predicts that people will balance if both sides of the market prefer more of a unequal traits through exchange.For example, characteristic in their mates,then neither side highly educated men from disadvantaged class www.analreviews.orgTrends and Variation in Asortative Mating 453

SO39CH22-Schwartz ARI 26 June 2013 15:20 people compete with one another for preferred mates. The search for a romantic partner has often been compared with the search for a job. Just as people have a reservation wage below which they will not take a job, they may also have a set of minimum qualifications they seek in a partner below which they will not form a match. People may update their preferences on the basis of their experience in the marriage market, and the complex interaction between people’s potentially shifting preferences and their constraints generates observed matching patterns. These ideas underlie many behavioral models of mate selection, although various simplifying assumptions are made (e.g., Burdett & Coles 1997, Logan et al. 2008, Wong 2003). They have primarily been applied to marriage but also apply to the search for other types of romantic partners (e.g., Choo & Siow 2006, Sahib & Gu 2002). Becker’s (1974, 1981) economic model of marriage is often evoked in studies of assortative mating because of its clear predictions about matching patterns. He predicted that like will marry like when traits are complements: characteristics such as lifestyle, attractiveness, and religion. Becker referred to this phenomenon as positive assortative mating. The terms endogamy and exogamy are also often used, respectively, for in- or out-group marriage (e.g., by religion, race/ethnicity, or nativity), and homogamy and heterogamy for marriages between people with similar or dissimilar traits (e.g., by years of schooling, income, or attractiveness). Other scholars have noted that homogamy can be generated by two combinations of preferences for mates: (a) Both sides of the market prefer partners with characteristics similar to their own (the matching hypothesis), or (b) both sides of the market prefer partners with more of a characteristic (the competition hypothesis). The matching hypothesis leads to homogamy because everyone is searching for someone similar to themselves. The competition hypothesis leads to homogamy because, if both sides of the market prefer more of a characteristic in their mates, then neither side wants to partner “down,” and everyone ends up with someone roughly similar to themselves. A small literature attempts to adjudicate between these two hypotheses (Dimaggio & Mohr 1985, Kalmijn 1994), with recent studies in economics using novel data from internet dating websites, speed dating experiments, and marriages among movie stars (Bruze 2011, Fisman et al. 2006, Hitsch et al. 2010). Overall, the evidence favors the matching hypothesis, but whether matching or competition prevails may vary by the trait in question, e.g., competition for income and matching on education, race/ethnicity, and religion (Hitsch et al. 2010). All these studies were conducted using data from the United States, however, and whether matching or competition characterizes the sorting process may vary by context. Becker predicted that negative assortative mating would occur for traits that are substitutes. Traits are substitutes when there are gains to specialization. Becker reasoned that, just as factory output can increase when individuals specialize rather than complete all parts of a process, the gains to marriage can be greater when members of a couple specialize, and individuals would thus sort into marriage to maximize these gains. Although specialization could occur on numerous household tasks (e.g., I’m better at tidying up, and my partner is better at washing dishes), Becker primarily focused on specialization in the broad categories of market work versus household work. As others have pointed out, Becker’s ideas are reminiscent of Parsons’s (1949) classic discussion of the conjugal family, in which the division of labor between husbands and wives serves to reduce competition between the sexes and promote marital stability. The gains to specialization led Becker to predict negative assortative mating on wages such that high-wage men (or women) marry low-wage women (or men). Exchange theory (Davis 1941, Merton 1941) also predicts negative assortative mating, but for different reasons. In the absence of homogamy, it predicts that people will balance unequal traits through exchange. For example, highly educated men from disadvantaged class www.annualreviews.org • Trends and Variation in Assortative Mating 453 Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

backgrounds can exchange their educational origin,race/ethnicity,religion)to achieved advantage to marry women from advantaged characteristics (e.g.,education,experience). class backgrounds with lower educational In the process,parents lose control over the attainments.The outcome of these matches is economic success of their children,and the negative assortative mating on both education influence of third parties (e.g.,families,com- and class,but it arises because people exchange munity members,the church,the state)over one advantage for another.Whether patterns mate selection wanes.Increasing geographic of exchange are evident in interracial marriages mobility also reduces the ability of parents in the United States has been the subject of and third parties to monitor and regulate lively debate (Gullickson Fu 2010;Kalmijn mate selection,but educational institutions 2010b:Rosenfeld2005,2010). and residential segregation remain important indirect methods of parental control (Blau Duncan 1967,Goode 1970 [1963]). TRENDS AND VARIATION: Assortative mating scholars have drawn on CAUSES these ideas-which can be loosely classified The ideas discussed above form the theoretical under the modernization theory umbrella-to basis for many studies of assortative mating,but predict an increase in sorting on achieved taken alone they are static.Theories of trends characteristics and a decrease in sorting on and variation in assortative mating require an ascribed characteristics (e.g.,Kalmijn 1991a,b; engine of change,be it variation in preferences, Rosenfeld 2006,2008;Rosenfeld Thomas in opportunities,or in some combination 2012).A variant of this hypothesis posits that of both.Although there is a small literature sorting on achieved characteristics follows an devoted to estimating the independent effects inverted U shape,first rising with the increased of preferences versus opportunities (e.g., importance of education but then falling as Abramitzky et al.2011,Choo Siow 2006, romantic love becomes the basis of mate Logan et al.2008),many of the ideas discussed selection (Smits et al.1998). below involve changes in both.In addition, Trends in assortative mating by religion feedback loops between preferences and and race/ethnicity are generally consistent with opportunities may render their independent predictions from modernization theory.With effects difficult to disentangle.Thus,while I use few exceptions,interracial and interreligious the concepts of preferences and opportunities marriages are vastly more common today than below,I do not dwell on these distinctions but in the past (e.g.,Kalmijn 1991a,Rosenfeld organize the discussion around the major sub- 2008).Trends in educational homogamy, stantive explanations for variation in assortative by contrast,are less consistent and more mating. contested.Several scholars have remarked that cross-national trends in educational homogamy appear to have no discernible pattern(Blossfeld Modernization and the Decline 2009,Kalmiin 1998,Smits Park 2009).In the of Third Party Control United States,most studies show a rise in edu- A constellation of interrelated ideas about mod- cational homogamy,but the results vary across ernization,economic development,urbaniza- the education distribution and are sensitive to tion,geographic mobility,secularization,and how spousal resemblance is measured (Hou the rise of individualism motivate a large Myles 2008,table 1;Rosenfeld 2008;Schwartz portion of theory about change and variation Mare 2005).Thus,at present,it is clear that in assortative mating.Although these ideas are cross-national trends in educational homogamy formulated in various ways,the basic argument do not neatly follow those predicted by mod- is that as societies develop,the basis of success ernization theory and that different or more shifts from ascribed characteristics (e.g.,social complex theories are needed,as advocated by 454 Sdrwartz

SO39CH22-Schwartz ARI 26 June 2013 15:20 backgrounds can exchange their educational advantage to marry women from advantaged class backgrounds with lower educational attainments. The outcome of these matches is negative assortative mating on both education and class, but it arises because people exchange one advantage for another. Whether patterns of exchange are evident in interracial marriages in the United States has been the subject of lively debate (Gullickson & Fu 2010; Kalmijn 2010b; Rosenfeld 2005, 2010). TRENDS AND VARIATION: CAUSES The ideas discussed above form the theoretical basis for many studies of assortative mating, but taken alone they are static. Theories of trends and variation in assortative mating require an engine of change, be it variation in preferences, in opportunities, or in some combination of both. Although there is a small literature devoted to estimating the independent effects of preferences versus opportunities (e.g., Abramitzky et al. 2011, Choo & Siow 2006, Logan et al. 2008), many of the ideas discussed below involve changes in both. In addition, feedback loops between preferences and opportunities may render their independent effects difficult to disentangle. Thus, while I use the concepts of preferences and opportunities below, I do not dwell on these distinctions but organize the discussion around the major substantive explanations for variation in assortative mating. Modernization and the Decline of Third Party Control A constellation of interrelated ideas about modernization, economic development, urbanization, geographic mobility, secularization, and the rise of individualism motivate a large portion of theory about change and variation in assortative mating. Although these ideas are formulated in various ways, the basic argument is that as societies develop, the basis of success shifts from ascribed characteristics (e.g., social origin, race/ethnicity, religion) to achieved characteristics (e.g., education, experience). In the process, parents lose control over the economic success of their children, and the influence of third parties (e.g., families, community members, the church, the state) over mate selection wanes. Increasing geographic mobility also reduces the ability of parents and third parties to monitor and regulate mate selection, but educational institutions and residential segregation remain important indirect methods of parental control (Blau & Duncan 1967, Goode 1970 [1963]). Assortative mating scholars have drawn on these ideas—which can be loosely classified under the modernization theory umbrella—to predict an increase in sorting on achieved characteristics and a decrease in sorting on ascribed characteristics (e.g., Kalmijn 1991a,b; Rosenfeld 2006, 2008; Rosenfeld & Thomas 2012). A variant of this hypothesis posits that sorting on achieved characteristics follows an inverted U shape, first rising with the increased importance of education but then falling as romantic love becomes the basis of mate selection (Smits et al. 1998). Trends in assortative mating by religion and race/ethnicity are generally consistent with predictions from modernization theory. With few exceptions, interracial and interreligious marriages are vastly more common today than in the past (e.g., Kalmijn 1991a, Rosenfeld 2008). Trends in educational homogamy, by contrast, are less consistent and more contested. Several scholars have remarked that cross-national trends in educational homogamy appear to have no discernible pattern (Blossfeld 2009, Kalmijn 1998, Smits & Park 2009). In the United States, most studies show a rise in educational homogamy, but the results vary across the education distribution and are sensitive to how spousal resemblance is measured (Hou & Myles 2008, table 1; Rosenfeld 2008; Schwartz & Mare 2005). Thus, at present, it is clear that cross-national trends in educational homogamy do not neatly follow those predicted by modernization theory and that different or more complex theories are needed, as advocated by 454 Schwartz Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

others (Smits Park 2009).A subset of the Economic Inequality ideas associated with modernization theory The economic inequality hypothesis represents has fared better in recent work on the decline an alternative to straight-line theories of mod- of third party control over mate selection and ernization and assimilation.It posits that the the rise of same-sex and interracial unions economic and social distance between groups is (Rosenfeld Kim 2005,Rosenfeld Thomas greater in times of high inequality,thereby re- 2012) ducing intermarriage.This hypothesis is quite general and can be applied to inequality across Assimilation many social groups,but it is most often applied Although some studies of trends in racial/ethnic to economic inequality by education(known as intermarriage draw on modernization theory, the returns to schooling)and to general lev- more common is the use of assimilation theory. els of income inequality across individuals or Like modernization theory,classical assimila- families (e.g,as measured by the Gini coeffi- Ajuo asn tion theory predicts an increase in racial/ethnic cient,Theil index,or coefficient of variation). intermarriage as the social boundaries between The mechanism linking inequality to assorta- euosiad groups blur through the process of assimilation tive mating is straightforward:When economic into a common culture(Gordon 1964,Park inequality across groups increases,people have Burgess 1921).Although long-run patterns of more to lose if they"marry down"(Fernandez racial/ethnic intermarriage fit squarely within et al.2005). predictions from both modernization and as- The inequality hypothesis is not new(Blau similation theory,an unprecedented decline in 1977,Rytina et al.1988,Smits et al.1998), Hispanic intermarriage between the mid-1990s but has received increased attention as a result 宝 and 2000s has sparked renewed interest among of several promising cross-national studies scholars of interracial/interethnic marriage in that have found higher odds of educational the link between preferences and opportunity homogamy in countries with greater income (e.g.,Lichter et al.2011;Qian Lichter 2007,inequality and returns to schooling(Dahan 2011).The leading explanation for Hispan-Gaviria 2001,Fernandez et al.2005,Torche ics'recent"retreat from intermarriage"is that 2010).Studies of trends within countries sup- -510c the large growth in the Hispanic immigrant port these findings:Educational homogamy population has increased opportunities for en-tends to move in tandem with the returns dogamy and may have also reinforced a shared to schooling (Han 2010,Heaton Mitchell cultural identity (Qian Lichter 2007,2011).2012).Related studies find that higher levels Aold The complement to these findings is that,while of male income inequality are associated with Hispanics in the United States have become delayed marriage among women (Gould increasingly endogamous,immigration out of Paserman 2003,Loughran 2002),potentially Mexico appears to have created a marriage mar-because whom one marries matters more in ket squeeze in which women living in areas of high inequality regimes than in low ones. Mexico with high levels of migration are more Whereas the inequality hypothesis is often likely to "marry down"in education than are framed as a shift in preferences stemming those living in areas with low migration(Choi from change in the cost of "marrying down,' Mare 2012).Broadly speaking,the variability of increases in income inequality are also asso- changes in intermarriage by race/ethnic group ciated with increased residential segregation supports more nuanced theories of assimilation (Reardon Bischoff 2011)and thus may reduce that are less unidirectional than classical assim-the opportunities for intermarriage.One way to ilation theory and follow multiple paths(e.g.,make progress toward identifying why inequal- Alba Nee 2003,Portes Zhou 1993,Qian ity and assortative mating are related would Lichter 2007). be to investigate the links between inequality, www.anmalreviews.org Trends and Variation in Asortative Mating 455

SO39CH22-Schwartz ARI 26 June 2013 15:20 others (Smits & Park 2009). A subset of the ideas associated with modernization theory has fared better in recent work on the decline of third party control over mate selection and the rise of same-sex and interracial unions (Rosenfeld & Kim 2005, Rosenfeld & Thomas 2012). Assimilation Although some studies of trends in racial/ethnic intermarriage draw on modernization theory, more common is the use of assimilation theory. Like modernization theory, classical assimilation theory predicts an increase in racial/ethnic intermarriage as the social boundaries between groups blur through the process of assimilation into a common culture (Gordon 1964, Park & Burgess 1921). Although long-run patterns of racial/ethnic intermarriage fit squarely within predictions from both modernization and assimilation theory, an unprecedented decline in Hispanic intermarriage between the mid-1990s and 2000s has sparked renewed interest among scholars of interracial/interethnic marriage in the link between preferences and opportunity (e.g., Lichter et al. 2011; Qian & Lichter 2007, 2011). The leading explanation for Hispanics’ recent “retreat from intermarriage” is that the large growth in the Hispanic immigrant population has increased opportunities for endogamy and may have also reinforced a shared cultural identity (Qian & Lichter 2007, 2011). The complement to these findings is that, while Hispanics in the United States have become increasingly endogamous, immigration out of Mexico appears to have created a marriage market squeeze in which women living in areas of Mexico with high levels of migration are more likely to “marry down” in education than are those living in areas with low migration (Choi & Mare 2012). Broadly speaking, the variability of changes in intermarriage by race/ethnic group supports more nuanced theories of assimilation that are less unidirectional than classical assimilation theory and follow multiple paths (e.g., Alba & Nee 2003, Portes & Zhou 1993, Qian & Lichter 2007). Economic Inequality The economic inequality hypothesis represents an alternative to straight-line theories of modernization and assimilation. It posits that the economic and social distance between groups is greater in times of high inequality, thereby reducing intermarriage. This hypothesis is quite general and can be applied to inequality across many social groups, but it is most often applied to economic inequality by education (known as the returns to schooling) and to general levels of income inequality across individuals or families (e.g., as measured by the Gini coeffi- cient, Theil index, or coefficient of variation). The mechanism linking inequality to assortative mating is straightforward: When economic inequality across groups increases, people have more to lose if they “marry down” (Fernandez ´ et al. 2005). The inequality hypothesis is not new (Blau 1977, Rytina et al. 1988, Smits et al. 1998), but has received increased attention as a result of several promising cross-national studies that have found higher odds of educational homogamy in countries with greater income inequality and returns to schooling (Dahan & Gaviria 2001, Fernandez et al. 2005, Torche ´ 2010). Studies of trends within countries support these findings: Educational homogamy tends to move in tandem with the returns to schooling (Han 2010, Heaton & Mitchell 2012). Related studies find that higher levels of male income inequality are associated with delayed marriage among women (Gould & Paserman 2003, Loughran 2002), potentially because whom one marries matters more in high inequality regimes than in low ones. Whereas the inequality hypothesis is often framed as a shift in preferences stemming from change in the cost of “marrying down,” increases in income inequality are also associated with increased residential segregation (Reardon & Bischoff 2011) and thus may reduce the opportunities for intermarriage. One way to make progress toward identifying why inequality and assortative mating are related would be to investigate the links between inequality, www.annualreviews.org • Trends and Variation in Assortative Mating 455 Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

segregation,and homogamy.Previous research literature on marriage is that,although Becker's has examined whether income segregation is as- (1974,1981)and Parsons's (1949)theories sociated with health and education outcomes, about gender specialization may once have but findings from these studies are mixed aptly characterized marital arrangements, (Lobmayer Wilkinson 2002,Mayer 2001, marriage is an increasingly egalitarian and Mayer Sarin 2005).Linking segregation to flexible institution(e.g.,Cherlin 2004;Gerson assortative mating seems quite promising given 2010;Oppenheimer 1994,1997).This argu- that relationship formation is strongly rooted ment underlies two main hypotheses about in space (McPherson et al.2001). how changes in gender inequality affect Despite some straightforward differences matching:one framed from the perspective of between the modernization and inequality hy-women choosing men,and the other from the potheses,few studies have empirically adjudi-perspective of men choosing women. cated between them.For instance,the two the- When women choose men,increases in ories diverge with respect to predicted trends women's socioeconomic status relative to in sorting on social origins.With increased de-men's should give women more freedom to velopment,modernization theory predicts a de-choose their mates on nonpecuniary grounds cline in sorting on social origins,but growing (Oppenheimer Lew 1995,Sweeney 2002). US wealth inequality since the 1970s (Wolff In other words,when choosing between "love 1998)suggests that the importance of sorting or money,"women can afford to choose love on social origins may have increased.Kalmijn (Fernandez et al.2005).This hypothesis is (1991b)found weak evidence of a decline in so- generally interpreted to imply less matching cial origin homogamy among couples married on socioeconomic resources,although it is between 1927 and 1961,but no study has ex-not clear why this should necessarily be the amined these trends since inequality began its case.For instance,because lifestyles,attitudes, steep rise.Additional studies of sorting on so- and beliefs are associated with class position cial origins would also be useful in addressing (Weeden Grusky 2012),it may be that whether increases in educational homogamy in increases in women's earnings facilitate a the United States reflect increased sorting on longer search,allowing women to find better class background,or whether men and women matches on numerous characteristics. sort on class background just as they always have From the perspective of men choosing even as educational homogamy has increased. women,as women's economic prospects grow Finally,interest in the inequality hypoth-and egalitarian marriage becomes more nor- esis has centered on variation in educational mative,men may begin competing for highly homogamy,but it could easily be extended to educated,high-earning women just as women racial/ethnic and religious endogamy.A fruit-have traditionally competed for high-earning ful avenue of future research would be to in-men (e.g.,England Farkas 1986,Sweeney vestigate the extent to which increases inin- Cancian 2004).The increasing similarity termarriage by race/ethnicity and religion can of men's and women's preferences drives be explained by changing economic differences up competition for high-status partners and between these groups. results in increased homogamy.Thus,unlike the usual interpretation from the perspective of women,this hypothesis implies that declining Gender Inequality gender inequality increases homogamy. The economic inequality hypothesis is not From the perspective of women choosing explicitly gendered,but changes in inequality men,the evidence on whether love trumps between men and women likely play an im-money as gender inequality decreases is mixed. portant role in assortative mating.A central Trend studies show that the importance of argument in the sociological and demographic men's earnings for marriage has not declined 456 Sdrwartz

SO39CH22-Schwartz ARI 26 June 2013 15:20 segregation, and homogamy. Previous research has examined whether income segregation is associated with health and education outcomes, but findings from these studies are mixed (Lobmayer & Wilkinson 2002, Mayer 2001, Mayer & Sarin 2005). Linking segregation to assortative mating seems quite promising given that relationship formation is strongly rooted in space (McPherson et al. 2001). Despite some straightforward differences between the modernization and inequality hypotheses, few studies have empirically adjudicated between them. For instance, the two theories diverge with respect to predicted trends in sorting on social origins. With increased development, modernization theory predicts a decline in sorting on social origins, but growing US wealth inequality since the 1970s (Wolff 1998) suggests that the importance of sorting on social origins may have increased. Kalmijn (1991b) found weak evidence of a decline in social origin homogamy among couples married between 1927 and 1961, but no study has examined these trends since inequality began its steep rise. Additional studies of sorting on social origins would also be useful in addressing whether increases in educational homogamy in the United States reflect increased sorting on class background, or whether men and women sort on class background just as they always have even as educational homogamy has increased. Finally, interest in the inequality hypothesis has centered on variation in educational homogamy, but it could easily be extended to racial/ethnic and religious endogamy. A fruitful avenue of future research would be to investigate the extent to which increases in intermarriage by race/ethnicity and religion can be explained by changing economic differences between these groups. Gender Inequality The economic inequality hypothesis is not explicitly gendered, but changes in inequality between men and women likely play an important role in assortative mating. A central argument in the sociological and demographic literature on marriage is that, although Becker’s (1974, 1981) and Parsons’s (1949) theories about gender specialization may once have aptly characterized marital arrangements, marriage is an increasingly egalitarian and flexible institution (e.g., Cherlin 2004; Gerson 2010; Oppenheimer 1994, 1997). This argument underlies two main hypotheses about how changes in gender inequality affect matching: one framed from the perspective of women choosing men, and the other from the perspective of men choosing women. When women choose men, increases in women’s socioeconomic status relative to men’s should give women more freedom to choose their mates on nonpecuniary grounds (Oppenheimer & Lew 1995, Sweeney 2002). In other words, when choosing between “love or money,” women can afford to choose love (Fernandez et al. 2005). This hypothesis is ´ generally interpreted to imply less matching on socioeconomic resources, although it is not clear why this should necessarily be the case. For instance, because lifestyles, attitudes, and beliefs are associated with class position (Weeden & Grusky 2012), it may be that increases in women’s earnings facilitate a longer search, allowing women to find better matches on numerous characteristics. From the perspective of men choosing women, as women’s economic prospects grow and egalitarian marriage becomes more normative, men may begin competing for highly educated, high-earning women just as women have traditionally competed for high-earning men (e.g., England & Farkas 1986, Sweeney & Cancian 2004). The increasing similarity of men’s and women’s preferences drives up competition for high-status partners and results in increased homogamy. Thus, unlike the usual interpretation from the perspective of women, this hypothesis implies that declining gender inequality increases homogamy. From the perspective of women choosing men, the evidence on whether love trumps money as gender inequality decreases is mixed. Trend studies show that the importance of men’s earnings for marriage has not declined 456 Schwartz Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

(Buss et al.2001,Sweeney 2002).But have much higher educational attainment than Fernandez et al.(2005)found that educational women do,unless a substantial fraction of the homogamy tends to be lower in countries population remains unmarried,men will tend where women are relatively economically to marry women with less education than them- advantaged,consistent with the notion that selves(a pattern called hypergamy).Men's ed- love trumps money with decreasing gender ucational advantage has declined substantially inequality.A key difference between Fernandez around the world,and in many countries,in- et al.'s(2005)study and others is that it controls cluding the United States,wives now have on for general levels of inequality across countries. average more education than their husbands(a It is possible that the effects of economic pattern called hypogamy)(Esteve et al.2012). inequality trump those of gender inequality An exception to this pattern is Japan,where but that once economic inequality is controlled highly educated women appear to be avoiding for,the effects of gender inequality can be marriage in the face of a decline in the rela- Ajuo asn observed.Future studies should be mindful of tive supply of highly educated men(Raymo these potentially offsetting effects. Iwasawa 2005). From the perspective of men choosing women,strong and growing evidence sug- gests that women's money has become more Changes in the Structure of Search important.There is a growing expectation Other scholars focus on changes in the struc- that women will contribute to their families' ture of searching for a mate in explaining varia- economic well-being (Gerson 2010,South tion in assortative mating.Aprominent hypoth- 1991),and these expectations are reflected in esisis Mare's(1991)time gap hypothesis,which 宝 men's attitudes and behaviors.Men with high argues that educational homogamy varies as a earnings potential are more likely than in the function ofthe timing of marriage and ofschool past to marry women with high wages(Sweeney completion.Because schools are educationally Cancian 2004).They also attach increasing homogeneous institutions,couples who meet in importance to women's financial prospects,school and marry shortly thereafter should be education,and intelligence in their search for more likely to be educationally homogamous mates(Buss et al.2001).These changes may be than those who meet in other places,such as due to the rise of egalitarian marriage,but they work,bars and nightclubs,or neighborhoods could also be spurred by the increased need for Indeed,Mare (1991)finds that part of the in- women's contributions arising from increased crease in educational homogamy in the United childrearing costs,declining male wages,rising States between 1940 and 1987 can be explained Aold standards of living,and increased uncertainty by the narrowing time gap between school com- in men's and women's economic trajectories pletion and marriage. (Oppenheimer 1988,Sweeney 2002). More recent trends in the time gap are Because relationships involve two people,inconsistent with trends in educational ho- it is of course possible for both hypotheses mogamy in the United States.The time gap to be correct-women could care less about has increased since the early 1980s as increases men's money,and men could care more about in schooling have fallen behind increases in age women's.How this plays out into observed at first marriage,but homogamy has not de- matching patterns is the result of a complex clined as would be expected.The inconsistency interplay between the preferences of men and between recent trends in the time gap and ed- women and the availability of potential mates.ucational homogamy in the United States may Indeed,the discussion above focuses on shifts explain why this hypothesis has received sur- in men's and women's preferences,but reduc-prisingly little empirical attention.An excep- tions in gender inequality also affect matching tion is cross-national research showing that,in through changes in opportunities.When men many countries,homogamy rises with time out www.analreviews.orgTrends and Variation in Asortative Mating 457

SO39CH22-Schwartz ARI 26 June 2013 15:20 (Buss et al. 2001, Sweeney 2002). But Fernandez et al. (2005) found that educational ´ homogamy tends to be lower in countries where women are relatively economically advantaged, consistent with the notion that love trumps money with decreasing gender inequality. A key difference between Fernandez ´ et al.’s (2005) study and others is that it controls for general levels of inequality across countries. It is possible that the effects of economic inequality trump those of gender inequality but that once economic inequality is controlled for, the effects of gender inequality can be observed. Future studies should be mindful of these potentially offsetting effects. From the perspective of men choosing women, strong and growing evidence suggests that women’s money has become more important. There is a growing expectation that women will contribute to their families’ economic well-being (Gerson 2010, South 1991), and these expectations are reflected in men’s attitudes and behaviors. Men with high earnings potential are more likely than in the past to marry women with high wages (Sweeney & Cancian 2004). They also attach increasing importance to women’s financial prospects, education, and intelligence in their search for mates (Buss et al. 2001). These changes may be due to the rise of egalitarian marriage, but they could also be spurred by the increased need for women’s contributions arising from increased childrearing costs, declining male wages, rising standards of living, and increased uncertainty in men’s and women’s economic trajectories (Oppenheimer 1988, Sweeney 2002). Because relationships involve two people, it is of course possible for both hypotheses to be correct—women could care less about men’s money, and men could care more about women’s. How this plays out into observed matching patterns is the result of a complex interplay between the preferences of men and women and the availability of potential mates. Indeed, the discussion above focuses on shifts in men’s and women’s preferences, but reductions in gender inequality also affect matching through changes in opportunities. When men have much higher educational attainment than women do, unless a substantial fraction of the population remains unmarried, men will tend to marry women with less education than themselves (a pattern called hypergamy). Men’s educational advantage has declined substantially around the world, and in many countries, including the United States, wives now have on average more education than their husbands (a pattern called hypogamy) (Esteve et al. 2012). An exception to this pattern is Japan, where highly educated women appear to be avoiding marriage in the face of a decline in the relative supply of highly educated men (Raymo & Iwasawa 2005). Changes in the Structure of Search Other scholars focus on changes in the structure of searching for a mate in explaining variation in assortative mating. A prominent hypothesis is Mare’s (1991) time gap hypothesis, which argues that educational homogamy varies as a function of the timing of marriage and of school completion. Because schools are educationally homogeneous institutions, couples who meet in school and marry shortly thereafter should be more likely to be educationally homogamous than those who meet in other places, such as work, bars and nightclubs, or neighborhoods. Indeed, Mare (1991) finds that part of the increase in educational homogamy in the United States between 1940 and 1987 can be explained by the narrowing time gap between school completion and marriage. More recent trends in the time gap are inconsistent with trends in educational homogamy in the United States. The time gap has increased since the early 1980s as increases in schooling have fallen behind increases in age at first marriage, but homogamy has not declined as would be expected. The inconsistency between recent trends in the time gap and educational homogamy in the United States may explain why this hypothesis has received surprisingly little empirical attention. An exception is cross-national research showing that, in many countries, homogamy rises with time out www.annualreviews.org • Trends and Variation in Assortative Mating 457 Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

of school (inconsistent with the time gap hy-networking websites may have lengthened the pothesis)but then falls(as predicted by the hy-reach of schools as marriage markets further pothesis)(Blossfeld Timm 2003).Another into adulthood (Ellison et al.2007).Social exception is an interesting study by Shafer networking websites may also facilitate the Qian (2010),which shows that the association expansion of social contacts along homophilous between the time gap and homogamy differs lines so that even partners who do not meet in by sex.As men spend more time out of school,school may match with someone who attended they are less likely to form educationally ho-a school with characteristics similar to those of mogamous unions,but women become more their own (Arum et al.2008).In addition,in- likely to avoid marriage altogether rather than ternet dating websites and social media reduce "marry down."This is consistent with inter-the costs of search,making it possible to find a net dating studies showing that women have a better and possibly more homogamous match. greater aversion to“marrying down'”in edu- By contrast,the dominant view in studies of cation than men do (Hitsch et al.2010).Black new technologies is that the internet promotes women,in particular,are more likely to remain diversity rather sameness because of increased unmarried than marry a man with less educa-contact with people from different backgrounds tion (Lichter et al.1995).Thus,the effect of and the reduced influence of third parties(e.g., the time gap may be more complex than origi- Hampton et al.2011,Rosenfeld Thomas nally outlined,varying by sex,race,and age.Fu-2012).Despite these arguments,what is per- ture research is needed to determine how these haps most striking about the internet dating complexities map onto variation in educational literature is how closely people's stated prefer- homogamy over time and place. ences for mates match what would be expected It is possible that the time gap hypothesis on the basis of observed matches in the gen- fits the meeting and marriage patterns of the eral population (Feliciano et al.2009,Hitsch 1940s through the 1980s but that this link has et al.2010).This suggests that matches formed become weaker with the changing structure of through the internet may not differ substan- search.With increased residential and occu-tially from those formed in other ways,but re- pational segregation by socioeconomic status search on how technology changes the nature (Hellerstein et al.2008,Reardon Bischoff of marriage markets and romantic relationships 2011)and reduced sex segregation at work is in its infancy. (Padavic Reskin 2002),young people may be nearly as likely to find a homogamous mate at a local bar or at work as they once were at school. Other Demographic Changes These processes may also work differently Demographic changes such as increasing by class.Recent evidence suggests that the cohabitation,rising divorce and remarriage, expansion of education has created a marriage and declining marriage rates may affect com- market mismatch for college students from less parisons of spousal resemblance across time advantaged backgrounds.Because college stu- and place(Schwartz Mare 2012).A common dents in general are still relatively advantaged, question is whether the rise of cohabitation can disadvantaged students who attend college may explain observed trends in assortative mating. face restricted partner markets and may also be This could occur if,for example,cohabita- hesitant to marry someone with less education tion weeds out heterogamous couples who upon returning home (Musick et al.2012). would have married in its absence(Blackwell Changes in dating technology may also Lichter 2000,2004).With respect to education, affect the structure of search.One possibility scholars have found little evidence of a weeding is that new technologies increase homogamy.mechanism-cohabitors who split up and The ability of individuals to keep in touch with those who marry are very similar with respect their school friends using cell phones and social to their educational resemblance (Goldstein 458 Sdrwartz

SO39CH22-Schwartz ARI 26 June 2013 15:20 of school (inconsistent with the time gap hypothesis) but then falls (as predicted by the hypothesis) (Blossfeld & Timm 2003). Another exception is an interesting study by Shafer & Qian (2010), which shows that the association between the time gap and homogamy differs by sex. As men spend more time out of school, they are less likely to form educationally homogamous unions, but women become more likely to avoid marriage altogether rather than “marry down.” This is consistent with internet dating studies showing that women have a greater aversion to “marrying down” in education than men do (Hitsch et al. 2010). Black women, in particular, are more likely to remain unmarried than marry a man with less education (Lichter et al. 1995). Thus, the effect of the time gap may be more complex than originally outlined, varying by sex, race, and age. Future research is needed to determine how these complexities map onto variation in educational homogamy over time and place. It is possible that the time gap hypothesis fits the meeting and marriage patterns of the 1940s through the 1980s but that this link has become weaker with the changing structure of search. With increased residential and occupational segregation by socioeconomic status (Hellerstein et al. 2008, Reardon & Bischoff 2011) and reduced sex segregation at work (Padavic & Reskin 2002), young people may be nearly as likely to find a homogamous mate at a local bar or at work as they once were at school. These processes may also work differently by class. Recent evidence suggests that the expansion of education has created a marriage market mismatch for college students from less advantaged backgrounds. Because college students in general are still relatively advantaged, disadvantaged students who attend college may face restricted partner markets and may also be hesitant to marry someone with less education upon returning home (Musick et al. 2012). Changes in dating technology may also affect the structure of search. One possibility is that new technologies increase homogamy. The ability of individuals to keep in touch with their school friends using cell phones and social networking websites may have lengthened the reach of schools as marriage markets further into adulthood (Ellison et al. 2007). Social networking websites may also facilitate the expansion of social contacts along homophilous lines so that even partners who do not meet in school may match with someone who attended a school with characteristics similar to those of their own (Arum et al. 2008). In addition, internet dating websites and social media reduce the costs of search, making it possible to find a better and possibly more homogamous match. By contrast, the dominant view in studies of new technologies is that the internet promotes diversity rather sameness because of increased contact with people from different backgrounds and the reduced influence of third parties (e.g., Hampton et al. 2011, Rosenfeld & Thomas 2012). Despite these arguments, what is perhaps most striking about the internet dating literature is how closely people’s stated preferences for mates match what would be expected on the basis of observed matches in the general population (Feliciano et al. 2009, Hitsch et al. 2010). This suggests that matches formed through the internet may not differ substantially from those formed in other ways, but research on how technology changes the nature of marriage markets and romantic relationships is in its infancy. Other Demographic Changes Demographic changes such as increasing cohabitation, rising divorce and remarriage, and declining marriage rates may affect comparisons of spousal resemblance across time and place (Schwartz & Mare 2012). A common question is whether the rise of cohabitation can explain observed trends in assortative mating. This could occur if, for example, cohabitation weeds out heterogamous couples who would have married in its absence (Blackwell & Lichter 2000, 2004).With respect to education, scholars have found little evidence of a weeding mechanism—cohabitors who split up and those who marry are very similar with respect to their educational resemblance (Goldstein 458 Schwartz Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

Harknett 2006,Sassler McNally 2003, tative mating trends to measure changes in the Schwartz 2010b).Given the lack of evidence strength of boundaries between social groups, for a weeding mechanism out of cohabitation, then addressing changes in selection into it is not surprising that estimates of trends in marriage is less relevant,but if one's interest is educational homogamy using pooled samples to understand the mechanisms through which of cohabitors and married couples are similar these patterns are generated,then differential to those using married couples alone (Qian selection into unions should be addressed. Preston 1993).There is more evidence for the weeding hypothesis with respect to interra- cial/interethnic and interreligious relationships TRENDS AND VARIATION: (Blackwell Lichter 2004,Joyner Kao 2005, CONSEQUENCES Wanget al.2006),but again,estimates of trends One of the main justifications for studying as- using pooled samples are similar to those using sortative mating patterns is that they may have Ajuo asn married couples alone (Qian Lichter 2007). profound consequences for inequality and pop- These findings suggest that increases in co- ulation change.Until recently,this hypothesis euosiad habitation have not substantially altered trends has received little empirical attention.Below, in assortative mating,at least in the United I review the literature on the consequences of States.Patterns of matching in cohabitation assortative mating for (a)inequality within gen- and marriage differ substantially across contexts erations,(b)inequality between generations, (Hamplova 2009),and the effects of cohabita- (c)long-run population change,and (d)rela- tion on assortative mating into marriage may tionship quality and dissolution. differ in turn.Similar to cohabitation,divorce 宝 and remarriage have small effects on the resem- blance between spouses in the United States, Inequality Within Generations where these events are relatively common,and Studies of the effects of assortative mating on therefore are unlikely to be primary expla- inequality within generations have focused on nations for large shifts over time or between spouses'socioeconomic resemblance.Because countries(Fu 2010,Schwartz Mare 2012). spouses share resources,societies in which high Most of the evidence about the effects earners marry other high earners and low earn- -510c of declining marriage rates on changes in ers marry other low earners will be more un- assortative mating comes from evidence on equal than those in which high earners marry cohabitation.This is reasonable given that low earners.To the extent that race/ethnicity 夏 the rise of cohabitation makes up for a large and religion are also associated with economic portion of the decline in marriage (Bumpass and cultural resources,assortative mating on et al.1991).But clearly,changes in selection these characteristics has implications for the into marriage cannot be completely captured distribution of resources across families and by the rise of cohabitation.For example,as households as well,but these implications have mentioned above,rising levels of inequality af-yet to be explored. fect how long women search for their marriage Studies of the impact of assortative mat- partners and potentially whether they marry at ing on inequality within generations can be all.Whether this induces women to substitute viewed as one piece of a larger literature cohabitation for marriage is an open question.on the effects of changes in family life on Some studies have used selection models to inequality.Related literature has found that examine differences in assortative mating steep increases in nonmarital childbearing, while controlling for differential selection into delayed marriage,and declining marriage unions (e.g.,Qian et al.2005),but whether rates have decreased the share of married- this is necessary depends on the question of couple households in the population.Because interest.If one's interest is to examine assor-single people tend to have lower household www.anmalreviews.orgTrends and Variation in Asortative Mating 459

SO39CH22-Schwartz ARI 26 June 2013 15:20 & Harknett 2006, Sassler & McNally 2003, Schwartz 2010b). Given the lack of evidence for a weeding mechanism out of cohabitation, it is not surprising that estimates of trends in educational homogamy using pooled samples of cohabitors and married couples are similar to those using married couples alone (Qian & Preston 1993). There is more evidence for the weeding hypothesis with respect to interracial/interethnic and interreligious relationships (Blackwell & Lichter 2004, Joyner & Kao 2005, Wang et al. 2006), but again, estimates of trends using pooled samples are similar to those using married couples alone (Qian & Lichter 2007). These findings suggest that increases in cohabitation have not substantially altered trends in assortative mating, at least in the United States. Patterns of matching in cohabitation and marriage differ substantially across contexts (Hamplova 2009), and the effects of cohabitation on assortative mating into marriage may differ in turn. Similar to cohabitation, divorce and remarriage have small effects on the resemblance between spouses in the United States, where these events are relatively common, and therefore are unlikely to be primary explanations for large shifts over time or between countries (Fu 2010, Schwartz & Mare 2012). Most of the evidence about the effects of declining marriage rates on changes in assortative mating comes from evidence on cohabitation. This is reasonable given that the rise of cohabitation makes up for a large portion of the decline in marriage (Bumpass et al. 1991). But clearly, changes in selection into marriage cannot be completely captured by the rise of cohabitation. For example, as mentioned above, rising levels of inequality affect how long women search for their marriage partners and potentially whether they marry at all. Whether this induces women to substitute cohabitation for marriage is an open question. Some studies have used selection models to examine differences in assortative mating while controlling for differential selection into unions (e.g., Qian et al. 2005), but whether this is necessary depends on the question of interest. If one’s interest is to examine assortative mating trends to measure changes in the strength of boundaries between social groups, then addressing changes in selection into marriage is less relevant, but if one’s interest is to understand the mechanisms through which these patterns are generated, then differential selection into unions should be addressed. TRENDS AND VARIATION: CONSEQUENCES One of the main justifications for studying assortative mating patterns is that they may have profound consequences for inequality and population change. Until recently, this hypothesis has received little empirical attention. Below, I review the literature on the consequences of assortative mating for (a) inequality within generations, (b) inequality between generations, (c) long-run population change, and (d ) relationship quality and dissolution. Inequality Within Generations Studies of the effects of assortative mating on inequality within generations have focused on spouses’ socioeconomic resemblance. Because spouses share resources, societies in which high earners marry other high earners and low earners marry other low earners will be more unequal than those in which high earners marry low earners. To the extent that race/ethnicity and religion are also associated with economic and cultural resources, assortative mating on these characteristics has implications for the distribution of resources across families and households as well, but these implications have yet to be explored. Studies of the impact of assortative mating on inequality within generations can be viewed as one piece of a larger literature on the effects of changes in family life on inequality. Related literature has found that steep increases in nonmarital childbearing, delayed marriage, and declining marriage rates have decreased the share of marriedcouple households in the population. Because single people tend to have lower household www.annualreviews.org • Trends and Variation in Assortative Mating 459 Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

incomes than married couples do,declines in not surprising given that homogamy increased married-couple families exacerbate inequality more for the most and least educated than for (McCall Percheski 2010,McLanahan those in the middle (Kalmijn 1991a,Rosenfeld Percheski 2008).The rising association 2008,Schwartz Mare 2005).These findings between husbands'and wives'earnings means suggest that educational homogamy may have that high-earning married-couple families are larger effects on inequality between high- additionally advantaged relative to singles. and low-earning couples than on inequality Estimates of the effects of the growing associa- between high-and middle-earning couples or tion between spouses'earnings on increases in low-and middle-earning couples.To date, inequality across married-couple families range studies of the effects of trends in educational from 17 to 51%,depending on the measures homogamy have used summary measures used and time period studied(Schwartz 2010a). of inequality,which could obscure variation The growing association between spouses' across the earnings distribution. earnings could be due to increases in assorta- A third explanation for the“no effects'”find- tive mating on socioeconomic characteristics, ing is that wives'education may not be highly to changes in the division of labor within mar- correlated with their earnings.As an extreme riage,or to some combination of both.As men- example,if all women exit the labor force tioned above,evidence suggests an increase in upon marriage,then the correlation between educational homogamy in the United States, wives'education and earnings would be zero, but whether this has translated into increased and increases in educational homogamy could inequality within generations is questionable.not affect earnings inequality.This extreme Two recent studies have found that increased example has not been far from the truth for educational homogamy has not contributed to much of US history.In 1950,only 21%of mar- increasedinequality in the United States(Breen ried white women were in the labor force.By Salazar 2011,Western et al.2008). 1980,the rate had risen to 49%-a substantial Although it is intuitive that increased edu- increase,but more than half of women still had cational homogamy should increase economic potential earnings that went unrealized(Goldin inequality across families because of the link be- 1990,table 2.1).Over this period,wives'labor tween education and earnings,there are several supply became more responsive to their own potential explanations for the "no effects"find- potential wages and less responsive to their ing.First,increases in educational homogamy husbands'wages(Goldin 2006).Because wives' may not have been large enough to produce siz- potential wages are shaped by their education, able shifts in inequality (Breen Salazar 2011, these findings suggest that a relatively weak Kremer 1997).A second,related possibility association between wives'education and earn- is that increases in educational homogamy ings may be responsible for the null findings. among some types of couples have been offset This explanation is also supported by the find- by declines among other types such that there ing that changes in educational homogamy have is little overall effect on inequality.Several had a larger effect on inequality in Denmark, scholars have pointed out that the correlation where asubstantially higher percentageofwives between spouses'educational attainments has work,than in the United States(Breen An- barely budged or,by some estimates,has even dersen 2012,Breen Salazar 2011).It would be declined in the United States(Breen Salazar interesting to extend this line of research by 2011,Kremer 1997,Rosenfeld 2008).How-testing how the effects of assortative mating on ever,models that describe trends in educational inequality vary depending on wives'labor force homogamy using association measures(such as participation and the strength of the associa- the correlation coefficient)do not fit observed tion between wives'education and earnings. trends as well as more complex models do It is possible that increases in the association (Mare 2000,Schwartz Mare 2005).This is between wives'education and earnings will ￥60 Sdrwartz

SO39CH22-Schwartz ARI 26 June 2013 15:20 incomes than married couples do, declines in married-couple families exacerbate inequality (McCall & Percheski 2010, McLanahan & Percheski 2008). The rising association between husbands’ and wives’ earnings means that high-earning married-couple families are additionally advantaged relative to singles. Estimates of the effects of the growing association between spouses’ earnings on increases in inequality across married-couple families range from 17 to 51%, depending on the measures used and time period studied (Schwartz 2010a). The growing association between spouses’ earnings could be due to increases in assortative mating on socioeconomic characteristics, to changes in the division of labor within marriage, or to some combination of both. As mentioned above, evidence suggests an increase in educational homogamy in the United States, but whether this has translated into increased inequality within generations is questionable. Two recent studies have found that increased educational homogamy has not contributed to increased inequality in the United States (Breen & Salazar 2011, Western et al. 2008). Although it is intuitive that increased educational homogamy should increase economic inequality across families because of the link between education and earnings, there are several potential explanations for the “no effects” finding. First, increases in educational homogamy may not have been large enough to produce sizable shifts in inequality (Breen & Salazar 2011, Kremer 1997). A second, related possibility is that increases in educational homogamy among some types of couples have been offset by declines among other types such that there is little overall effect on inequality. Several scholars have pointed out that the correlation between spouses’ educational attainments has barely budged or, by some estimates, has even declined in the United States (Breen & Salazar 2011, Kremer 1997, Rosenfeld 2008). However, models that describe trends in educational homogamy using association measures (such as the correlation coefficient) do not fit observed trends as well as more complex models do (Mare 2000, Schwartz & Mare 2005). This is not surprising given that homogamy increased more for the most and least educated than for those in the middle (Kalmijn 1991a, Rosenfeld 2008, Schwartz & Mare 2005). These findings suggest that educational homogamy may have larger effects on inequality between highand low-earning couples than on inequality between high- and middle-earning couples or low- and middle-earning couples. To date, studies of the effects of trends in educational homogamy have used summary measures of inequality, which could obscure variation across the earnings distribution. A third explanation for the “no effects” finding is that wives’ education may not be highly correlated with their earnings. As an extreme example, if all women exit the labor force upon marriage, then the correlation between wives’ education and earnings would be zero, and increases in educational homogamy could not affect earnings inequality. This extreme example has not been far from the truth for much of US history. In 1950, only 21% of married white women were in the labor force. By 1980, the rate had risen to 49%—a substantial increase, but more than half of women still had potential earnings that went unrealized (Goldin 1990, table 2.1). Over this period, wives’ labor supply became more responsive to their own potential wages and less responsive to their husbands’ wages (Goldin 2006). Because wives’ potential wages are shaped by their education, these findings suggest that a relatively weak association between wives’ education and earnings may be responsible for the null findings. This explanation is also supported by the finding that changes in educational homogamy have had a larger effect on inequality in Denmark, where a substantially higher percentage of wives work, than in the United States (Breen & Andersen 2012, Breen & Salazar 2011). It would be interesting to extend this line of research by testing how the effects of assortative mating on inequality vary depending on wives’ labor force participation and the strength of the association between wives’ education and earnings. It is possible that increases in the association between wives’ education and earnings will 460 Schwartz Annu. Rev. Sociol. 2013.39:451-470. Downloaded from www.annualreviews.org Access provided by Shanghai Jiaotong University on 09/27/16. For personal use only

点击进入文档下载页（PDF格式）

共23页，试读已结束，阅读完整版请下载

点击下载（PDF格式）

浏览记录