Qian and Oian:The gender divide in urban China:Singlehood and assortative mating by age and education wife are in the same age groups.As a result,we overestimate the odds of age homogamy. We employ crossings models.Table 1 presents the crossings parameters in detail. Suppose that intermarriage is a process of crossing barriers of different levels.Under crossings models each barrier is determined by which two adjacent levels it separates. For instance,the barrier between less than high school and high school is v,the barrier between high school and vocational college is v2,and so on (Hout 1983).Thus, crossings models can reveal which two educational levels are serious barriers to intermarriage (Mare 1991).Parameters in Table I indicate log odds of intermarriage across two adjacent educational levels relative to log odds of homogamy,controlling for marginal distributions of husband's and wife's education.Prospective spouses with a greater distance in education must cross more barriers to get married,i.e.,the log odds of marriage for couples across several educational levels are the sum of the crossings parameters separating husbands'and wives'educational levels (Schwartz and Mare 2005).Crossings models are gender-symmetrical,and assume that the difficulty of crossing each educational barrier is the same,no matter whether the husband or the wife has more education. Table 1: Parameters for crossings effects on educational assortative marriage Wives'Educational Attainment Husbands'Educational Less than Vocational High school College or Attainment high school college above Less than high school 1 V h+2 V1+2+3 High school V1 之 V2 2+V3 Vocational college 1+V2 2 1 g College or above 1+V2+V3 2+V3 V3 Note:Table is adapted from Schwartz and Mare (2005) When gender asymmetry and crossings models are included,the model is as follows: IogN脚=BO+∑BMA XMA+∑BWA XWA+∑BME XME+∑BWE XWE +BMAME XMAXME+∑WAWE XWAXWE+∑BHE XHE (2) +∑B哈XAE+∑BHAXHA+∑BAA XAA+∑BCE XCE where Nw denotes the expected number of marriages between men aged i with education k and women aged j with education /Bo is an intercept and other Bs denote 1346 http://www.demographic-research.orgQian and Qian: The gender divide in urban China: Singlehood and assortative mating by age and education 1346 http://www.demographic-research.org wife are in the same age groups. As a result, we overestimate the odds of age homogamy. We employ crossings models. Table 1 presents the crossings parameters in detail. Suppose that intermarriage is a process of crossing barriers of different levels. Under crossings models each barrier is determined by which two adjacent levels it separates. For instance, the barrier between less than high school and high school is ν1, the barrier between high school and vocational college is ν2, and so on (Hout 1983). Thus, crossings models can reveal which two educational levels are serious barriers to intermarriage (Mare 1991). Parameters in Table 1 indicate log odds of intermarriage across two adjacent educational levels relative to log odds of homogamy, controlling for marginal distributions of husband‟s and wife‟s education. Prospective spouses with a greater distance in education must cross more barriers to get married, i.e., the log odds of marriage for couples across several educational levels are the sum of the crossings parameters separating husbands‟ and wives‟ educational levels (Schwartz and Mare 2005). Crossings models are gender-symmetrical, and assume that the difficulty of crossing each educational barrier is the same, no matter whether the husband or the wife has more education. Table 1: Parameters for crossings effects on educational assortative marriage Wives’ Educational Attainment Husbands’ Educational Attainment Less than high school High school Vocational college College or above Less than high school 1 ν1 ν1 + ν2 ν1 + ν2 + ν3 High school ν1 1 ν2 ν2 + ν3 Vocational college ν1 + ν2 ν2 1 ν3 College or above ν1 + ν2 + ν3 ν2 + ν3 ν3 1 Note: Table is adapted from Schwartz and Mare (2005). When gender asymmetry and crossings models are included, the model is as follows: logNijkl = +∑ + ∑ + ∑ + ∑ + ∑ + ∑ + ∑ (2) + ∑ + ∑ + ∑ + ∑ where Nijkl denotes the expected number of marriages between men aged i with education k and women aged j with education l. is an intercept and other s denote