Research on assortative mating over the past three decades shows a tendency of individuals to marry others with similar characteristics. Positive marital sorting increases income inequality and is positively correlated with wage inequality. Even though marriage choice has important implications for inequality and its persistence, little is known about how marital sorting patterns arise. The positive correlation between spouses’ characteristics reflects two forces in the marriage market: demand and supply. Investigating mating preferences in the process of union formation is central to understand marriage market dynamics leading to social inequality. This dissertation project focuses on the process of partner choice. In my first chapter, I investigate educational homogamy by union type (i.e. cohabitation and marriage) in one Latin American country: Mexico. I test two hypotheses that explain differences in educational homogamy between marriage and cohabitation. The “winnowing” hypothesis assumes that people become more selective as they move from dating to cohabiting to marriage, hence it predicts higher homogamy in marriages than in cohabitations. The “looser bond” hypothesis assumes that cohabitation is a living arrangement chosen by individuals with more egalitarian values seeking a relationship lacking long-term commitment. The “looser bond” hypothesis conceives education as an indicator of potential labor market success and individual autonomy and predicts higher educational homogamy in cohabitation than in marriage. I use data from three waves of the Mexican Family Life Survey (MxFLS) and follow the stock-and-flow framework proposed by Schwartz (2010). This model investigate educational homogamy by comparing patterns of association by union type (i.e. cohabitation vs. marriage), but it goes further and takes advantage of longitudinal data to investigate some of the implications of these hypotheses on selective union dissolution. To examine differences in educational homogamy, I rely on log-linear models to investigate spousal resemblance across union type and across transition status. I find no support for the “looser bond” hypothesis and strong support for the “winnowing” hypothesis. Results are also consistent with the cultural matching hypothesis and with recent economic theories that suggest the main difference between cohabitation and marriage is the strength of the commitment. Given that cohabitation is an institution of shorter duration than marriage, education becomes more salient for couples opting for marriage and this explains higher homogamy rates in marriages than in cohabitation. Finally, the evidence shown in this chapter also supports the idea that cohabitation markets, in Mexico, are less structured by education compared to marriage markets.